What's New

AI branding priced, and fits on a phone

Generating a theme from your description now costs 1 token in Basic; it used to run for nothing. Advanced pricing is unchanged. The participant and prize rows now wrap on a phone, so the Add buttons stay on screen.

Maturity chart fits on a phone

On a phone the maturity chart numbers its five levels and names your current level underneath, instead of pushing the page sideways.

Settings labels match your plan

Duration, follow-up questions and brain teasers no longer carry an Advanced badge: Basic has them. The comparison table now shows all six roles and difficulty 1-5 in Basic. The interview toolbar wraps on a phone so End stays on screen.

Table shows your real limits

The comparison table now shows 25 jobs and up to 5 search links in Basic, which is what the tool allows. The sort row wraps on a phone instead of running off the screen.

Table and setup list match the quiz

The comparison table now shows every difficulty and 4 to 6 answer options in Basic. PDF and PowerPoint reports are no longer listed as Advanced-only on the setup screen; every plan has them.

Notes counter shows the analysis limit

The notes box shows a counter out of 3,000 characters and tells you when notes beyond that will be left out of the analysis. The teleprompter still shows your notes in full.

Batch hint names your plan limit

The upload hint now says batches take up to 10 receipts on Pro and 5 on other plans, matching what the tool accepts.

Cost assumptions in plain words

The cost assumptions now read as labels such as Developer rate and Build approach, not raw field names. The comparison table no longer lists PDF export for the demo; it needs an account.

Clear Use Case message, tabs fit on phones

Starting without a Use Case now says what is missing and moves you to the field. The results tabs wrap on a phone instead of pushing the page sideways. The comparison table now shows that refinement and version history are included in Basic.

Comparison table names the right model

The comparison table now names the AI model Advanced mode actually runs.

Right model listed; tidier on phones

The comparison table now names the AI model Advanced mode actually runs, and the context row wraps on narrow screens.

Model description matches the picker

The Advanced mode note now points to the model shown below it instead of naming a different one.

Clear message when text is too long

Text too long for a QR code now tells you the limit and shows a character count, and the old code can no longer be downloaded by mistake. The comparison table no longer lists an SVG download.

Comparison table matches the demo

The comparison table now shows what the demo already includes: sound, editing, fullscreen, winner history and PDF/PowerPoint export. Prize images work on every plan.

Correct AI model and batch sizes listed

The comparison table now names the model Advanced mode really uses, and shows batch uploads as up to 10 on Pro and up to 5 on other plans.

WebM output works again

Converting to WebM failed every time with "The video engine stopped". WebM is now made with VP8, which plays wherever WebM does. Files can be a little larger than before.

WebM clips work again

WebM clips failed every time with "The video engine stopped". They are now made with VP8, which plays wherever WebM does. GIF and MP4 are unchanged.

Transparent masks work

The transparent mask export failed, and even when it ran, the area outside your regions came out solid instead of see-through. It now exports a WebM where everything outside your regions is transparent.

Non-Latin names kept, plan shown honestly

Exported dictionaries keep Ukrainian and other non-Latin dataset names. If your plan cannot be checked, the tool says so and stays unlocked for the visit instead of treating you as Free.

Refinements read your whole scenario

Refining a decision used to send only the first part of a long scenario, without saying so. It now uses the whole scenario, and text that is genuinely too long gets a clear "too long" message instead.

Errors show on the results screen

If an export, chat or revision fails while you are looking at your results, the error now appears on that screen. Before, it went to the input screen you had already left, so the button seemed to do nothing.

Profile fields say when they are too long

Profile fields such as desired role, location and career goals now stop at their limit as you type. Anything over it gets a clear message naming the field and its limit.

No fallacy planted where you cannot answer it

The opponent no longer plants a fallacy in the final round, where you have no turn left to call it. In the demo, a whole debate now uses one demo run, and a debate that has already ended says so instead of spending another.

Cyrillic renders in Text to PDF

Text to PDF now renders Cyrillic instead of blank boxes. HTML to Markdown or text drops scripts and styles and keeps tables, a bad file in a batch is named and skipped, and non-Latin file names are kept.

Clear text limits, and no false downgrade

Title, topic, notes and follow-up questions stop at their limit as you type, and anything longer gets a clear "too long" message instead of being cut off without a word. A failed plan check no longer drops a paying user to the free features.

Word counts are right in every alphabet

Word counts are now correct for Cyrillic, Chinese and Japanese text and for words with curly apostrophes. In the demo, letting it ask you a few questions before writing is free, up to three times a day, and uses none of your demo runs.

Follow-ups see the whole analysis

Follow-up questions used to see only the first part of a long analysis. They now see all of it, so answers can draw on every project and recommendation.

Length warning only where it applies

The red export-length warning now shows only to Free users, who have the length limit. The mask file now records every region you drew.

Palettes give you every colour you asked for

A palette could come back one colour short, or with two swatches that were nearly the same. You now get the number of distinct colours you asked for, whenever the image has that many.

Paid plans are recognised again

The generator no longer treats every user as free, so paid features unlock for paying users. The billing choice stays on the toggle you set.

Step tests say why they failed

Testing a single step now shows the real reason when it fails, such as text that is too long or cannot be used, instead of a bare "Test failed". A prompt or step that is too long is refused with a clear message.

A failed extraction says so

An extraction that fails now shows an error instead of downloading an empty file. In a batch, a bad file is named and skipped and the rest still download. Non-Latin file names are kept.

Valid XML and complete CSV

XML output is now always valid, and CSV output includes every column and row. Downloads keep Ukrainian and other non-Latin file names.

Honest AVIF, and clear batch results

AVIF is offered only in browsers that can create it. A failed conversion shows an error instead of an empty file, a bad file in a batch is named and skipped, and non-Latin file names are kept.

Topic and description limits you can see

The topic and description boxes now stop at their limit as you type, and a request over it gets a plain message naming the field and its limit.

A failed conversion says so

A conversion that fails now shows an error instead of downloading an empty file. Batches go one file at a time: a bad file is named and skipped, and the rest still download. Non-Latin file names are kept.

Cleaning no longer collapses a file to one row

Removing duplicates could collapse a whole file to a single row. Every row that should stay now stays. If your plan cannot be checked, the tool says "Checking your plan" and stays unlocked for the visit instead of showing Free.

Merge adds the files you pick

Merging PDFs now adds the files you chose; before, it could miss them. Downloads keep Ukrainian and other non-Latin file names, and a failed plan check no longer shows you as Free.

Decimal commas and whole cents

You can type a decimal comma, such as 6,5, and it is read as 6.5. Loan schedules now round every payment to whole cents, and inputs that are not numbers get a clear message.

Catches IPv6, spaced IBANs and credentials

The redactor now catches IPv6 addresses, IBANs written with spaces, and passwords, keys and tokens left in text.

The demo says what a story costs

In the demo, adding Story to your session now tells you it uses one of your demo runs for today, so the count going down is not a surprise.

A failed clip says so

A clip or GIF that fails to render now shows an error instead of downloading an empty file, and Ukrainian and other non-Latin file names are kept.

Game words play without the wait

Flashcards, Multiple Choice, Listening, Typed Recall and Scramble now get the next word's pronunciation ready while you work on the current one, so a session without the Learn step no longer pauses on each card.

Auto-Detect names the contract type

With Contract Type left on Auto-Detect, the result now shows the type we found, such as "Mutual Non-Disclosure Agreement", instead of the word "auto".

A blocked microphone is explained until you fix it

If your browser blocks the microphone, the reason now stays on the recording screen, whatever panes you have open, with a Try again button for after you allow access.

The demo is clear about what it can do

The demo writes from the material you paste. Online research, checking against your sources and checking every claim are shown but switched off there, with a note that they are part of the full tool.

All 16 animation styles in the demo, draws free

The demo now says what it already offered: all 16 reveal styles, up to 20 entrants and 5 prizes. Running a draw no longer uses one of your daily demo runs.

One demo consultation counts as one run

In the demo, starting a schema consultation uses one of your daily runs. Answering its follow-up questions and generating the schema are part of that same run.

Progress bar stays sensible

Some videos could make the progress bar show an impossible percentage. It now holds steady and names the step it is on, and a failed engine download explains itself instead of saying "Failed to fetch".

Progress bar stays sensible

Some videos could make the progress bar show an impossible percentage. It now holds steady and names the step it is on, and a failed engine download explains itself instead of saying "Failed to fetch".

Progress bar stays sensible

Some videos could make the progress bar show an impossible percentage. It now holds steady and names the step it is on, and a failed engine download explains itself instead of saying "Failed to fetch".

Progress bar stays sensible

Videos recorded in a browser could make the progress bar show a large negative percentage. It now holds steady and names the step it is on. If the video engine cannot download, you get a clear message instead of "Failed to fetch".

Question limit shown correctly

The settings now say what you actually get: up to 15 questions per session when signed in, 5 in the demo. They previously listed a lower limit of 8 for Basic, which was never applied.

PDF uploads work in the demo

The demo now reads the text in your PDF, so an uploaded contract is analyzed instead of being refused. Scanned PDFs and photos get a clear note before you submit; the full tool still reads those. PDFs with no file type are recognised too.

The demo review no longer crashes

In the demo, asking the AI to review the sample workbook now lists every recommendation, starting with the first, instead of stopping on an error screen.

Clips and GIFs render again

Making a clip or GIF no longer gets stuck while the video engine loads, and a failed load can be retried without reloading the page.

MP4 downloads work again

Downloading a recording as MP4 works again. Before this fix it failed and offered the WebM file instead.

Extraction starts again

Extracting audio no longer gets stuck while the audio engine loads, and a failed load can be retried without reloading the page.

Conversions start again

Converting a video no longer gets stuck while the video engine loads, and a failed load can be retried without reloading the page.

Exports no longer stall at startup

Cropping and stitching no longer get stuck on "Initializing FFmpeg". If the video engine cannot start, you now see a clear message and can try again.

Long question sets and MP4 downloads work

Asking for up to 15 questions with follow-ups no longer fails partway through, and downloading a practice recording as MP4 works again instead of falling back to WebM.

Advanced screening no longer errors

Advanced screening with some AI models could fail with a server error even though the analysis had finished. Those results now come through.

Professional estimates finish reliably

Detailed Professional estimates no longer stop partway through on large jobs. Each step now gets the time it needs, and a slow step retries within the time left instead of failing the whole estimate.

Every document you upload is read, not just five

Attaching several supporting documents at once used to stop after five with a rate-limit notice, and the rest were dropped without being read. All of them now load. A long document also arrives in full, with a note on the chip if it had to be trimmed.

Longer resumes reach the analysis whole

A resume longer than the analysis limit used to be shortened on upload with nothing to say so. The whole document now comes through, and if it is still too long for the analysis you are told how much was trimmed.

Every document you upload is read, not just five

Attaching several documents at once used to stop after five with a rate-limit notice, and the rest were dropped without being read. All of them now load. A long resume or posting also arrives in full, with a note on the chip if it had to be trimmed.

Your recording is there even when we cannot score you

If speech recognition missed your answers, the screen that says so now shows your recording too — it was always saved, just never offered there. It stays when you start over, until you record another interview or delete it.

Backgrounds show their real price, and we ask before spending

Each AI background now shows what it actually costs before you generate one, and asks once per price rather than promising a rate we do not charge. If the price cannot be read, we say so instead of guessing. The usage bar keeps a running total for your visit.

See the price before you start, and what this visit has cost

The cost of a session is on screen before it begins, and the usage bar now keeps a running total for your visit — "This session: 1 token" — with a Start over link to begin counting again. Runs you start in another tab count too.

See the price before you start, and what this visit has cost

The cost of a debate is on screen before it opens, and the usage bar now keeps a running total for your visit — "This session: 1 token" — with a Start over link to begin counting again. Runs you start in another tab count too.

See the price before you generate, and what this visit has cost

Quiz and RPG generation show what they will charge before you run them, and the usage bar now keeps a running total for your visit — "This session: 2 tokens" — with a Start over link to begin counting again.

Fixes ask before they spend, at the price you are shown

Quick fixes and bulk apply now ask first, with the exact price in the question, and bulk apply asks once for the whole batch instead of charging fix by fix. The price quoted is the price charged. The usage bar keeps a running total for your visit.

The mic check shows how loud your room is

Next to the mic-check verdict you now see a Room number, 0 to 100: the steady background sound on its own, read once the check has listened for a second. The bars still show your voice above it. A quiet room reads low; a fan or traffic reads higher.

The mic check shows how loud your room is

Next to the mic-check verdict in setup you now see a Room number, 0 to 100: the steady background sound on its own, read once the check has listened for a second. The bars still show your voice above it. A quiet room reads low; a fan or traffic reads higher.

Copy confirms only after a real copy

Copy buttons now say "Copied" only once the text is on your clipboard, and tell you if your browser blocked the copy.

Copy tells you when it could not copy

Copying a suggested reply now reports a blocked clipboard instead of failing silently, so you know whether the text made it to your clipboard.

Copy output tells you when it could not copy

Copying the converted output now reports a blocked clipboard with a way forward, instead of failing silently.

Copying the winner announcement reports failures

Copying the winner announcement now tells you if your browser blocked the copy, instead of failing quietly.

Every Copy button confirms honestly

Copying a prompt, variant, negative prompt, or system prompt now confirms only after the text is on your clipboard, and tells you if your browser blocked the copy.

Share links confirm when copied

Copying a quiz link, challenge link, template link, or RPG join link now confirms once the link is on your clipboard, and tells you if your browser blocked it.

Copy tells you when it could not copy

Copying your plan summary now reports a blocked clipboard instead of failing quietly, so you always know whether the text made it to your clipboard.

Copying the QR image reports failures

Copying your QR code as an image now tells you when the browser cannot do it and suggests Download instead, rather than failing quietly.

Copy text tells you when it could not copy

Copying extracted text now reports a blocked clipboard with a way forward, instead of failing silently.

Copying a message reports failures

Copying a message now tells you if your browser blocked the copy, instead of failing quietly.

Copy confirms only after a real copy

Copy buttons now say "Copied" only once the text is on your clipboard, and tell you if your browser blocked the copy.

Post copies report failures

Copying a post, all posts, or an email now tells you if your browser blocked the copy, instead of failing quietly. "Copied" appears only after the copy succeeds.

Phrase copies confirm honestly

Copying a phrase or brief used to say "Copied" before the copy had happened. It now confirms only after the text is on your clipboard, and tells you if your browser blocked it.

Copy buttons confirm the copy

Copying suggested negotiation language now shows "Copied" once the text is on your clipboard, and tells you if your browser blocked the copy so you can select and copy it by hand.

Copying a share link reports failures

Copying your story's share link now tells you if your browser blocked the copy, instead of failing quietly.

Exports included on every plan

CSV, Excel, PDF and PowerPoint exports of your process chart now download on every plan, including Free.

Exports included on every plan

The analysis export now downloads on every plan, including Free.

Exports included on every plan

The session report export now downloads on every plan, including Free.

Exports included on every plan

Excel, PDF and PowerPoint exports of your portfolio review now download on every plan, including Free.

Exports included on every plan

The audit exports (CSV, Excel and PDF) now download on every plan, including Free. The password and cookie downloads stay with the Pro clean-up features. The demo still has no exports.

Exports included on every plan

The Excel clause table and the PDF report now download on every plan, including Free.

Exports included on every plan

CSV, Excel and PDF exports of your extracted invoices now download on every plan, including Free.

Exports included on every plan

Excel, PDF and PowerPoint exports of your journey plan now download on every plan, including Free. Nothing else about the exports changed.

Exports included on every plan

CSV and PDF exports now download on every plan, including Free. The formatted resume export (PDF/Word) uses AI, so it stays with Basic and Pro; on a Free account the buttons show a Basic mark instead of an error.

Exports included on every plan

JSON, Markdown and PDF exports now download on every plan, including Free. Markdown always downloads as a file now; the Copy button beside it still copies the text.

Exports included on every plan

CSV, Excel, PDF, SQL and Visio exports of your schema now download on every plan, including Free.

Exports included on every plan

Word, Excel, PDF and PowerPoint exports of your proposal now download on every plan, including Free. Basic accounts no longer hit an error on Excel or PowerPoint.

Exports included on every plan

Excel and PDF exports of your screening results now download on every plan, including Free.

Exports included on every plan

CSV, PDF and PowerPoint exports and the RPG narrative export now download on every plan, including Free. The demo still has no exports.

Exports included on every plan

Every export (client brief, Word, Excel, PowerPoint, ZIP and both calendar files) now shows in either mode on every plan, including Free. Add to Calendar on the Calendar tab and Download All on the Events tab are there for everyone.

Exports included on every plan

The Excel workbook is no longer Advanced-only: Excel, Markdown and PDF exports download in either mode on every plan, including Free.

Exports included on every plan

JSON and Mermaid downloads are no longer Advanced-only: all four formats (PDF, Markdown, JSON, Mermaid) download in either mode on every plan.

Exports included on every plan

Excel, PDF and PowerPoint exports of your estimate now download on every plan, including Free.

Every file you choose gets a clear answer

Choose an entrant list, prize picture, logo, background or sound and a note by the button names the file, its size and what we read from it — or why it was refused (wrong type, empty, over the limit). It stays until you dismiss it rather than vanishing after a few seconds.

Every file you choose gets a clear answer

Choose a background picture and a note by the button names the file, its size and what we read from it — or why it was refused (wrong type, empty, over the limit). It stays until you dismiss it rather than vanishing after a few seconds.

Every file you choose gets a clear answer

Choose a bookmarks export and a note by the button names the file, its size and what we read from it — or why it was refused (wrong type, empty, over the limit). It stays until you dismiss it rather than vanishing after a few seconds.

Every file you choose gets a clear answer

Choose an attachment and a note by the button names the file, its size and what we read from it — or why it was refused (wrong type, empty, over the limit). It stays until you dismiss it rather than vanishing after a few seconds.

Every file you choose gets a clear answer

Choose an STL model and a note by the button names the file, its size and what we read from it — or why it was refused (wrong type, empty, over the limit). It stays until you dismiss it rather than vanishing after a few seconds.

Every file you choose gets a clear answer

Choose a job posting file and a note by the button names the file, its size and what we read from it — or why it was refused (wrong type, empty, over the limit). It stays until you dismiss it rather than vanishing after a few seconds.

Every file you choose gets a clear answer

Choose a steps CSV and a note by the button names the file, its size and what we read from it — or why it was refused (wrong type, empty, over the limit). It stays until you dismiss it rather than vanishing after a few seconds.

Every file you choose gets a clear answer

Choose a source file or a shape and a note by the button names the file, its size and what we read from it — or why it was refused (wrong type, empty, over the limit). It stays until you dismiss it rather than vanishing after a few seconds.

Every file you choose gets a clear answer

Choose an existing plan file and a note by the button names the file, its size and what we read from it — or why it was refused (wrong type, empty, over the limit). It stays until you dismiss it rather than vanishing after a few seconds.

Every file you choose gets a clear answer

Choose a proposal to import and a note by the button names the file, its size and what we read from it — or why it was refused (wrong type, empty, over the limit). It stays until you dismiss it rather than vanishing after a few seconds.

Every file you choose gets a clear answer

Choose a CSV or JSON dataset and a note by the button names the file, its size and what we read from it — or why it was refused (wrong type, empty, over the limit). It stays until you dismiss it rather than vanishing after a few seconds.

Every file you choose gets a clear answer

Choose a resume file and a note by the button names the file, its size and what we read from it — or why it was refused (wrong type, empty, over the limit). It stays until you dismiss it rather than vanishing after a few seconds.

Every file you choose gets a clear answer

Choose a job posting or a custom avatar and a note by the button names the file, its size and what we read from it — or why it was refused (wrong type, empty, over the limit). It stays until you dismiss it rather than vanishing after a few seconds.

Every file you choose gets a clear answer

Choose a logo and a note by the button names the file, its size and what we read from it — or why it was refused (wrong type, empty, over the limit). It stays until you dismiss it rather than vanishing after a few seconds.

Every file you choose gets a clear answer

Choose an existing schema file and a note by the button names the file, its size and what we read from it — or why it was refused (wrong type, empty, over the limit). It stays until you dismiss it rather than vanishing after a few seconds.

Every file you choose gets a clear answer

Choose a data file and a note by the button names the file, its size and what we read from it — or why it was refused (wrong type, empty, over the limit). It stays until you dismiss it rather than vanishing after a few seconds.

Every file you choose gets a clear answer

Choose a plan or estimate to import and a note by the button names the file, its size and what we read from it — or why it was refused (wrong type, empty, over the limit). It stays until you dismiss it rather than vanishing after a few seconds.

The meter shows your voice, not the room

Steady background sound — a fan, a hum, traffic — now sits near zero and the bars rise on your voice. If your browser cannot transcribe speech (Firefox, for one) we say so before you record; if it stays silent while you talk, we tell you within seconds and name the language set.

We say when your browser cannot hear you

If your browser has no speech recognition (Firefox, for one) you are told on setup and on each question, and answers with no transcript are not scored or charged. If recognition stays silent while you speak, a notice appears within seconds and names the language set.

Your recording has sound and a real length; PDF job postings read as text

The full-interview video download now carries your voice, the player shows the true length and lets you scrub, and the caption says what the file is. A PDF or Word job posting arrives as text; a file we can't read says so under the field. Speaking bars sit clear of the caption.

A résumé we can't read is named, not attached

If a PDF or Word file can't be read — a scan with no text, a file over the limit — the profile now says which file and why, instead of attaching the file's raw code. The first file you attach is labelled Resume even if an earlier pick was refused.

Emoji now appear in your PDF exports

Emoji in your content — like 🚀 or ✅ — now print in exported PDFs instead of being dropped. They appear in a single ink colour.

Emoji now appear in your PDF exports

Emoji in your content — like 🚀 or ✅ — now print in exported PDFs instead of being dropped. They appear in a single ink colour.

Emoji now appear in your PDF exports

Emoji in your content — like 🚀 or ✅ — now print in exported PDFs instead of being dropped. They appear in a single ink colour.

Emoji now appear in your PDF exports

Emoji in your content — like 🚀 or ✅ — now print in exported PDFs instead of being dropped. They appear in a single ink colour.

Emoji now appear in your PDF exports

Emoji in your content — like 🚀 or ✅ — now print in exported PDFs instead of being dropped. They appear in a single ink colour.

Emoji now appear in your PDF exports

Emoji in your content — like 🚀 or ✅ — now print in exported PDFs instead of being dropped. They appear in a single ink colour.

Emoji now appear in your PDF exports

Emoji in your content — like 🚀 or ✅ — now print in exported PDFs instead of being dropped. They appear in a single ink colour.

Emoji now appear in your PDF exports

Emoji in your content — like 🚀 or ✅ — now print in exported PDFs instead of being dropped. They appear in a single ink colour.

Emoji now appear in your PDF exports

Emoji in your content — like 🚀 or ✅ — now print in exported PDFs instead of being dropped. They appear in a single ink colour.

Emoji now appear in your PDF exports

Emoji in your content — like 🚀 or ✅ — now print in exported PDFs instead of being dropped. They appear in a single ink colour.

Emoji now appear in your PDF exports

Emoji in your content — like 🚀 or ✅ — now print in exported PDFs instead of being dropped. They appear in a single ink colour.

Emoji now appear in your PDF exports

Emoji in your content — like 🚀 or ✅ — now print in exported PDFs instead of being dropped. They appear in a single ink colour.

Emoji now appear in your PDF exports

Emoji in your content — like 🚀 or ✅ — now print in exported PDFs instead of being dropped. They appear in a single ink colour.

Emoji now appear in your PDF exports

Emoji in your content — like 🚀 or ✅ — now print in exported PDFs instead of being dropped. They appear in a single ink colour.

Emoji now appear in your PDF exports

Emoji in your content — like 🚀 or ✅ — now print in exported PDFs instead of being dropped. They appear in a single ink colour.

See when a file import worked — or why it did not

Uploading a file to fill the form now shows the file by name and a clear result: how many fields it filled, or why it could not be read. The message stays until you dismiss it instead of flashing past, and a scanned PDF now tells you to try uploading it as an image.

Typed Recall now accepts synonyms and verb forms

In the type-the-meaning round, a correct answer written in different words — a true synonym, or a verb's matching aspect form in a language like Ukrainian — now counts, while a genuinely wrong word still does not. Newly built word sets carry these alternatives.

Pronunciations you've heard once now play instantly

The first time anyone hears a word, we save the audio — so every later play, for you or another learner, is instant instead of waiting on a fresh recording. Your Learn card also readies its word in advance, so the very first tap plays right away.

Reporting a wrong card now confirms — and won't double-report

Flagging a translation as wrong gave no confirmation, and the same card could be reported repeatedly. Now you get a clear "thanks — we're taking that one out of your deck" on the card, it leaves your deck as you move on, and can't be reported twice.

The previous word no longer plays over the next card

Tap Next while a word was still being pronounced and the sound could carry onto the next card. Moving on now stops the previous clip first — in every practice game and in the Learn walkthrough — so each card starts clean.

Every dollar figure is one the tool checked

Monthly and weekly amounts are now always added up into a yearly total, and the tool shows only figures it can trace to the numbers you entered. If it cannot check one against your inputs, it leaves it out rather than show a number that is not backed by your scenario.

Accents heard better, and a lost turn is never silent

Practise a language that isn't your first and Converse now tells the transcriber which language you speak, so it expects your accent specifically. And when a spoken turn can't be made out, you're told clearly and the cursor jumps to the type box.

Pick up a conversation where you left off

Leave a conversation and come back within 48 hours and Converse now offers to resume it — the same scene, and the transcript already on screen, so your partner carries on instead of starting over. Leaving still ends and refunds the unused minute; resuming starts a fresh one.

Coaching feedback in your own language

When you practise a language different from your own, the optional coaching notes on what you said now come back in your own language - not the one you are practising. The suggested replies and the example rewording still come in the language you are learning, where you want them.

Documents come out in the language you wrote in

Write your brief and notes in any language and Document Studio now writes the whole document — title, headings and body — in that language, instead of defaulting to English. Want it in a different language than your notes? Pick one from the new language menu.

Your argument score now rewards a genuinely strong turn

We recalibrated how each turn is scored so the number tracks real differences in your arguments. A clear, well-supported point that answers the other side now earns the higher score it deserves — and you can watch it climb as your debating sharpens over a session.

Exports warn instead of saving a blank file

If your browser can't render a scene into a video, ZIP, or slide export, you'll now see a clear message and a fix to try (the Premium render, or Chrome) instead of quietly downloading a blank or black file.

Chart category labels show in full

On bar and before/after comparison charts, long category names used to get trimmed with an ellipsis in your video and slide exports. They now show in full so every label stays readable.

Cover backgrounds stay on the cover in PDF and web exports

A cover-page background used to repeat on every page of a PDF export, and cover the whole document in the web (HTML) export. It now stays on the cover page, matching the Word file. Backgrounds set for every page, or for a section, are unaffected.

We tell you when a long résumé is trimmed

If your résumé is longer than the limit, the analyzer now trims it to fit and shows a clear note, instead of quietly using only the first part. Advanced mode allows more source text.

We tell you when a long résumé is trimmed

If your résumé is longer than the limit, the coach now trims it to fit and shows a clear note before your interview, instead of silently using only the beginning. Advanced mode allows more source text.

Your debate PDF stays readable on every page

The exported debate now uses its dark background on every page, not just the cover, so your arguments no longer appear as white text on a white page further into the document.

Your finished video downloads as a file

Downloading your rendered 1080p video now saves it to your device with a proper file name, instead of opening and playing inside the browser tab.

Form errors clear as soon as you fix them

A validation message on the situation box used to stay highlighted until you reloaded the page, even after you had started fixing it. It now clears the moment you edit the field.

More resume and job text on Advanced, clearer trimming

Advanced mode now accepts more of your resume and job posting (up to 32,000 characters each). If what you paste or upload runs over the limit, we now tell you it was trimmed, instead of silently using only the first part.

More source text on Advanced, and clearer trimming

Advanced mode now accepts up to 32,000 characters per source (was 12,000). If your sources run over the limit, Document Studio now tells you exactly what was shortened or left out, instead of quietly using only part of them.

Fullscreen draws scroll when there is a lot to show

Running a draw in fullscreen now scrolls when the winner details and controls do not all fit, so the Next Prize and View All Results buttons are always reachable. Also fixed the Fullscreen button doing nothing on some iPhones.

Unused minutes come back to you

When you finish a conversation, any whole minutes you paid for but never spoke into are returned to your balance. Converse buys your time a minute ahead so the audio never stalls; stop early and the minutes you did not reach come straight back.

The pause holds steady in a noisy room

The timer that pauses when you go quiet now holds up better around background noise. A fan, traffic or other voices no longer keep the clock running while you are not speaking, so a busy room will not quietly spend the time you were saving for talking.

The Compose button is clear in light mode

The Compose Story button and the input tabs could show faint text on a light background. They are now clearly legible in both light and dark themes.

Warnings are easy to read in light mode

Caution messages — like the note that only part of a very large paste was scanned — were hard to read on a light background. They now use a high-contrast amber so you won't miss them.

Press Enter to add the next line item

In the service, description and price fields, pressing Enter on the last row now adds a new line item — so you can build a quote from the keyboard without reaching for the mouse.

Fairer comparisons across monthly and yearly costs

When your options mix monthly, weekly and one-time costs, the advice now compares them on the same yearly footing before recommending — and if it picks the pricier option, it says why.

Games fund from your daily tokens, shared quizzes open

RPG adventures now draw on your daily token allowance before your purchased balance, so a game can start on daily tokens alone, and the daily counter updates as you play. Shared quiz links also open reliably for everyone you send them to.

CSV exports open cleanly in Excel

CSV files you convert now open in desktop Excel with accented and non-English characters intact, across both single and batch exports.

Exports open cleanly in Excel

Your cleaned CSV now opens in desktop Excel with accented and non-English characters intact — no more garbled text when you open the download.

Long entries wrap instead of overflowing

Long words and links in your analysis now wrap onto the next line instead of stretching past the edge, so your results stay readable on narrow screens.

PDF uploads are read again

Uploading a PDF résumé or job posting works again — the file reader that had stopped loading now runs, so your document is parsed as soon as you add it.

Share links work while you're signed in

Saving a story and getting a share link now works every time you're signed in — the "authentication required" message no longer appears when you click Share.

Your folder names survive the upload

Folders from a browser's second area - Chrome's Other bookmarks, Firefox's Bookmarks Menu - used to collapse to the top level, and exports with no toolbar wrapper lost every folder. They are kept now, and same-named folders in different places stay separate.

Organise your bookmarks your way

You choose how much reorganising happens, how many folders and how deep, and what to do when a site and its pages are both saved. It reads how you already organise and suggests. A new panel shows exactly what changed — and nothing is deleted unless you say so.

The duplicate count no longer includes what you removed

If you removed bookmarks during the review step, those removals were being added to the "Duplicates found" total on your results card and in the PDF audit. That number now counts duplicates only, so it always matches the list shown beside it.

Older http:// bookmarks are no longer called broken

Link checking marked every plain http:// bookmark as broken, and broken links default to Remove. Now we check the secure version of the address instead: if it answers, the bookmark is kept. If we still cannot check it, it is marked "couldn't verify" and left exactly where it is.

Links with several settings now survive the cleanup

Web addresses carrying more than one setting - the part after the "?" - could come back slightly altered after a cleanup, and the same bookmark saved in two browsers was not always seen as the same link. Both are fixed, so merging libraries matches far more of them.

The summary now tells you how many are real errors

The line at the top of your results now breaks the need-attention count down, so you can see how many of those are outright errors rather than warnings. A workbook with none reads exactly as it did before - nothing is added.

Both follow-ups arrive, and the interviewer keeps its voice

A second follow-up used to cancel the first, so only one ever reached you. And adding a follow-up cleared the pre-loaded interviewer audio, dropping the rest of the session to your browser's voice - and, with a persona, stopping the remaining videos.

Downloads match your plan, and empty plans aren't delivered

Every section the AI returns is now checked before it reaches you. Details shown on screen no longer go missing from your download, and stray placeholder text is gone. If a run returns nothing usable, you get a clear message and a one-click retry instead of an empty file.

Bookmarks with no name always get one

Bookmarks saved as "Untitled" or with no title at all sometimes came back exactly as they went in. Every one of them is now given a readable name taken from the page itself or its web address, and each rename is listed in the audit so you can see what changed.

The summary at the top now matches the list below it

The "healthy / need attention" line could disagree with the components listed underneath it. It counted cells the tool had already cleared, missed flagged hyperlinks, linked images, Power Query steps and orphan named ranges, and called a sheet healthy while marking it amber.

Stakeholder groups and model names you can trust

On Advanced runs, the Stakeholders tab now says when a group heading was written by the model that combines the analyses rather than drawn from your own analysis. Source model names are checked against the models that actually ran, so a model that never ran is no longer listed.

Mitigations and amplifiers now in your PDF

Your PDF download now lists the mitigations and amplifiers for each effect alongside its stakeholders - the same three lists you already see on screen and in the Markdown download. Advanced runs were producing all three; the PDF was leaving two of them out.

Refinements keep the rest of your result

Refining your optimized prompt used to replace the entire result, so parts of it could go missing. Now only what you asked about changes, the rest is kept, and a note names what was updated. If a refinement is cut off, nothing changes and we say so.

Refinements no longer drop parts of your plan

Asking to change one section used to replace your whole plan with whatever came back, so other sections could quietly disappear. Now only the sections you asked about change, everything else is kept, and a note tells you what was updated.

Auto-structure from a description now works

The button that turns a written process description into structured steps returned an error every time. It now reads your description and builds the step list, and you can edit every step afterwards as usual.

Import & Pre-fill now works, and says what it could not use

Import & Pre-fill returned an error every time you used it. It now reads your plan and fills the form in, matching your industry, size, frameworks and scenarios to the choices we offer. Rather than guess, it leaves alone anything it cannot place and names what did not match.

Agreement labels you can read and trust

On advanced multi-model runs the agreement count and source models could attach to the wrong effect, and the agreement badge could show an impossible figure. Both fixed. Details written by the model that combines the analyses are now labelled as such.

Chain step notes each get their own line

In the Markdown download, a chain step's input, output and handoff notes now appear as three separate lines rather than running together into one.

Analysis details read as separate lines

In the Markdown download, the details at the top of your analysis — when it was generated, the category, your tier and how many degrees were traced — each sit on their own line instead of running together into a single paragraph.

Cleaner recovery plan downloads

Severity and impact now sit on their own lines in the Markdown and PDF downloads, and playbook steps show their owner and timing cleanly. Unusual plan content no longer leaves stray placeholder text in the file.

Readable downloads and accurate confidence scores

Your markdown download now reads properly - each effect's details sit on separate lines instead of running together in one block. Key insights now appear in both the markdown and PDF downloads. Confidence percentages are right again, and the timestamp names its time zone.

Advanced mode now covers the follow-up questions too

On Advanced, the coach used its faster model for the extra questions it asks about your answers, while the rest of the session used the stronger one. Follow-ups now come from the same model as everything else, so they dig in the way your other questions do. Cost is unchanged.

Follow-up questions that follow your answer

The interviewer sometimes asked a follow-up about your previous answer rather than the one you had just given, and never asked one after the first question. Follow-ups now come from what you actually said. A skipped or unheard answer no longer triggers one.

A zero error count is not always a clean file

If a workbook was exported by another program rather than saved in Excel, its formula results are not stored yet, so an error inside one cannot be seen. We now say that instead of reporting a clean file. Open it in Excel, save, and upload it again for the full picture.

A job that changes nothing costs nothing

If a surgery finds nothing to change, you are not charged for it — and the panel now says so instead of promising a refund. Files that fail inside a batch come off the price of that run rather than being credited back afterwards.

Manual edits and cell edits are free

Editing a component's source, or a cell in the grid, still cost 1 credit even though manual edits moved to the free tier. They are free now. The price shown on the Manual Edit panel is also read from the same place the charge comes from, so the two cannot disagree again.

Deep analyses no longer get cut off

The most detailed Advanced runs — five degrees of effects reviewed by several models — could take long enough to be cut short before they finished. They now have the time they need, so a full-depth analysis returns its result instead of an error.

No more blank pages in your PDF

Exported PDFs no longer end with a page carrying only the footer. The report is the same, in fewer pages, and the footer sits at the bottom of a real page instead of on one of its own.

No more blank pages in your PDF

Campaign PDFs and briefs no longer carry pages holding only the footer. The same content now runs to about half the pages, with the footer at the bottom of a real page.

No more blank pages in your PDF

Exported PDFs no longer end with a page carrying only the footer. The report is the same, in fewer pages, and the footer sits at the bottom of a real page instead of on one of its own.

No more blank pages in your PDF

Quiz and story PDFs no longer end with a page carrying only the footer. The same content, in fewer pages, with the footer at the bottom of a real page.

No more blank pages in your PDF

Exported PDFs no longer end with a page carrying only the footer. The report is the same, in fewer pages, and the footer sits at the bottom of a real page instead of on one of its own.

Interview questions print as a section, not a page each

Interview questions now print as a proper section instead of one page per character when they arrive in an unexpected format, and questions with no category are grouped under General rather than a heading reading undefined.

No more blank pages in your PDF

Exported PDFs no longer end with a page carrying only the footer. The report is the same, in about half the pages, and the footer sits at the bottom of a real page instead of on one of its own.

No more blank pages in your PDF

Exported proposals no longer carry a page holding only the footer, so the document reads straight through and the footer sits at the bottom of a real page.

No more blank pages in your PDF

Exported PDFs no longer end with a page carrying only the footer. The report is the same, in about half the pages, and the footer sits at the bottom of a real page instead of on one of its own.

No more blank pages in your PDF

Exported PDFs no longer end with a page carrying only the footer. The report is the same, in fewer pages, and the footer sits at the bottom of a real page instead of on one of its own.

No more blank pages in your PDF

Exported PDFs no longer end with a page carrying only the footer. The estimate is the same, in fewer pages, and the footer sits at the bottom of a real page instead of on one of its own.

No more blank pages in your PDF

Exported PDFs no longer end with a page carrying only the footer. The report is the same, in fewer pages, and the footer sits at the bottom of a real page instead of on one of its own.

No more blank pages in your PDF

Exported PDFs no longer end with a page carrying only the footer. The report is the same, in about half the pages, and the footer sits at the bottom of a real page instead of on one of its own.

Long invoices print properly

Invoices with a long tax breakdown or a full set of confidence scores no longer spread a single line across each page. The same detail fits on a few full pages, and a field that arrives in an unexpected format is skipped rather than printed as rows of stray characters.

PDF and Excel downloads work again

Exporting your portfolio analysis to PDF or Excel had been failing since the tool launched - the button returned an error every time. Both now download the complete report again, with your organization name and budget included. PowerPoint was unaffected.

Errors are now found anywhere on the sheet

We only checked a corner of each sheet — the first 50 rows and 26 columns — so an error further down was never reported. Every cell is checked now; on one real workbook that surfaced 72 errors we never mentioned. Long lists are shortened, and we name any sheet we could not read.

Bulk Repair no longer refuses its own list

On a workbook with a lot of problems, "Repair everything" stopped with "Too many targets" and did nothing. It now repairs a full pass of them, spread across the kinds it found, and tells you how many are left so you can run it again.

Headings starting with # are no longer called errors

A cell like "#1 Priority" or "#hashtag" was flagged as a formula that had failed, purely because its text began with a hash. Only real Excel error values are reported now, so your own column headings stay out of the issue list and out of the error count.

Filled-down columns are read as formulas again

A column you filled down was being read as typed-in numbers, so a broken cell in one was never offered a repair. It is now recognised as a formula. We also stop calling a workbook password-protected when it isn't, and now say which part of each sheet was checked.

Cell errors now say what is actually wrong

Not every red cell is a broken formula. We now separate formulas that are genuinely faulty from ones waiting on a workbook we cannot reach, ones failing only because another cell does, and #N/A put there on purpose. Bulk Repair no longer offers to rewrite working cells.

Right page numbers on longer PDFs

A PDF that ran to several pages carried its footer on the last page only, and numbered every export "Page 1 of 1" however long it was. Longer analyses now show the footer and the correct page numbers throughout.

Longer PDF exports no longer fail

Exporting a prompt long enough to run onto a second page returned an error instead of a file. Every export now downloads, with the footer and page numbers on each page. Shorter one-page exports were unaffected.

Résumé PDFs upload properly again

Uploading a résumé as a PDF — on the search form or in your saved profile — had been failing since mid-August, and the match could end up built from the raw contents of the file. PDFs are read properly now. Word and plain-text uploads were never affected.

Résumé PDFs upload properly again

Attaching a résumé as a PDF had been failing since mid-August, and the interview could end up built from the raw contents of the file rather than your résumé. PDFs are read properly now. Word and plain-text attachments were never affected.

Résumé PDFs upload properly again

Uploading a résumé as a PDF had been failing since mid-August, whatever the file. PDFs are read properly now, and a file we genuinely cannot open says so instead of the upload failing with no explanation. Word and plain-text résumés were never affected.

PowerPoint downloads work again

Exporting to PowerPoint had been failing for about a week and showed only a generic error. It now downloads a complete deck again. Your content and your plan were never affected, and nothing you saved was lost. Only the download itself was broken.

PowerPoint downloads work again

Exporting your draw results to PowerPoint had been failing for about a week and showed only a generic error. It now downloads a complete deck again. Your draws and your plan were never affected, and nothing you saved was lost. Only the download itself was broken.

PowerPoint downloads work again

Exporting to PowerPoint had been failing for about a week and showed only a generic error. It now downloads a complete deck again. Your content and your plan were never affected, and nothing you saved was lost. Only the download itself was broken.

PowerPoint downloads work again

Exporting to PowerPoint had been failing for about a week and showed only a generic error. It now downloads a complete deck again. Your content and your plan were never affected, and nothing you saved was lost. Only the download itself was broken.

PowerPoint downloads work again

Exporting to PowerPoint had been failing for about a week and showed only a generic error. It now downloads a complete deck again. Your content and your plan were never affected, and nothing you saved was lost. Only the download itself was broken.

PowerPoint downloads work again

Exporting to PowerPoint had been failing for about a week and showed only a generic error. It now downloads a complete deck again. Your content and your plan were never affected, and nothing you saved was lost. Only the download itself was broken.

PowerPoint downloads work again

Exporting to PowerPoint had been failing for about a week and showed only a generic error. It now downloads a complete deck again. Your content and your plan were never affected, and nothing you saved was lost. Only the download itself was broken.

PowerPoint downloads work again

Exporting to PowerPoint had been failing for about a week and showed only a generic error. It now downloads a complete deck again. Your content and your plan were never affected, and nothing you saved was lost. Only the download itself was broken.

Breaking a link keeps your numbers

Removing an external link now leaves the value the cell last showed, the way Excel's own Break Links does. A cell whose value had never been calculated used to be emptied silently; it now reads #REF!, and the summary names every cell that happened to.

Describe a fix in your own words

After a workbook is scanned you get a short results screen listing what was found. Say what you want changed in plain English, or pick one of the named common jobs, and it runs. Each thing the scan can fix comes with a one-click offer.

The workbook tree now counts and names things correctly

A workbook with 24 external links listed 48 of them, half pointing at nothing. Each link is now named after the workbook it actually reaches for, instead of an internal file name, so you can tell which one is stale. Charts are counted once rather than once per styling file.

Refine now works on your recovery plan

Asking for a change on the results screen returned an error instead of revising your plan. Refinement now reaches the plan you generated, so feedback like "add a ransomware scenario" updates it. Available on Basic and Pro.

Refine now works on your optimized prompt

Asking for a refinement on the results screen returned an error every time instead of updating your prompt. Refinement now reaches the result you generated, so feedback like "make it shorter" or "add examples" produces a new version. Available on Basic and Pro.

Job links now read the posting, not the page around it

Paste a job URL and the tool reads the employer's own structured listing - the title, company, location and pay as fields - instead of whatever text it could scrape off the page. Postings that used to come back thin or empty now come back whole.

Fix several components in one pass

Tick any components in the tree and act on them together, with one plain-English instruction for the whole group. Ask for a workbook-wide change and the review's suggestions come back as ticks you can adjust first — nothing is proposed or charged until you confirm.

Some free actions were charging credits - fixed

Reviewing a whole workbook, previewing bulk changes, comparing two versions and rendering a chart preview are meant to be free, and now they cost nothing. One listed edit type that could never actually run has been removed, so it can no longer be suggested or charged for.

Keep your own network's links when cleaning a workbook

Strip external links, data connections, hyperlinks and linked images while keeping the ones on hosts you name, such as your SharePoint, intranet or file servers. Everything removed is listed afterwards, and a link the rule cannot judge is kept rather than quietly dropped.

World maps now match the country names you type

Typing "United States" now colours the United States. Everyday names — USA, UK, Czech Republic, Ivory Coast and more — are matched to the map for you. And a country your data has no figure for is now shaded a distinct grey, instead of looking like the lowest value on the chart.

World maps now show the whole world

The map chart drew only the countries you had data for, floating with no map behind them — and in the downloaded video it drew no map at all. Maps now render the full basemap, with countries that have no figure in a neutral grey so they are not read as a measured zero.

Pictograph charts now draw again

The pictograph chart — the one that repeats an icon per unit — was not drawing at all; a story that chose it showed an empty placeholder. It now renders everywhere, including the Premium video, and an icon name the writer gets slightly wrong is matched to the closest real icon.

Save your adventurers and bring them back for free

Keep up to 24 characters in your library. Bring saved ones into any new run and their portraits cost nothing, because they are already drawn -- a party of five with three saved is 7 tokens instead of 12. Pick them on the setup screen and the price updates before you pay.

See what keeps getting past you, and keep the debate

Your scorecard now shows which fallacies beat you across every debate, worst first, with the move that answers each. One you have never been shown says so rather than scoring zero. Download any finished debate as a PDF or Word file - both sides, every round, each fallacy marked.

Hear your opponent, and call it out loud

Your opponent now reads every reply aloud in their own voice, and you can mute it at any time - muting stops it being generated, not just played. The "call it" box takes dictation too, so you can name the fallacy the moment you hear it. The transcript is unchanged either way.

Every percentage in your scenario gets worked out

The advisor was working out some percentages from your scenario and quietly skipping others, like a 5% match on a stated salary. It now works out each one itself. Where a percentage could belong to two different amounts, it says nothing rather than guess.

Hear a partner before you choose one

Every partner card now has a play button. You hear their real opening line in their own voice, so you can pick by how someone sounds instead of by a name. It costs nothing and starts instantly - the clips are recorded, not generated when you press.

The clock stops when nobody is talking

Go quiet and the timer pauses - your remaining time is held, not spent. It starts again the moment you speak or type, or you can press Pick it back up. Leave it longer than half an hour and the conversation closes, with nothing charged for the wait.

A fallacy reference, and every charge shown in context

Open the fallacy reference at any point in a debate: all 26, what each one sounds like, and the sentence that answers it. Your scorecard now shows every charge inside the whole argument it was made in, with the fallacy marked - your own and the opponent's.

Dictation tells you which part is not working

The microphone button now recovers when the browser's speech service drops mid-sentence, instead of quietly giving up for the rest of the recording. And if no words come through, it checks your microphone separately and tells you whether the problem is your mic or the browser.

Converse has a card, and it opens

Converse now has its own artwork in the tools list, and its page opens the practice room instead of showing an empty panel. The page tells you a session is metered before you start one.

Nineteen situations, and feedback on what you said

Choose who you talk to and the situation - 21 partners across 19 scenarios, each in its own setting. Set the language you are learning and the one you think in. With feedback on, each thing you say gets a note on your wording, including when nothing needs changing.

Practise speaking out loud with a partner

Pick a partner tied to what you are working on - interviews, travel, or workplace English - and talk with them out loud. They answer in their own voice, follow what you actually said, and remember where you left off. Charged a minute at a time from your daily tokens.

Body-language frames are checked before they are read

If a captured frame is not a readable image, or is unusually large, the analysis now says so straight away and names which frame, instead of failing part-way through. Normal recordings are unaffected.

Body-language frames are checked before they are read

If a captured frame is not a readable image, or is unusually large, the analysis now says so straight away instead of failing part-way through. Normal recordings are unaffected.

Debate an opponent that argues badly on purpose

Argue any motion against an opponent that takes the other side and plants named logical fallacies on purpose. Call one out as you reply and find out straight away if you were right. The scorecard shows what you caught, what got past you, and the fallacies in your own arguments.

Retries that can actually finish

If a request to the AI runs long, the automatic retry now starts with a fresh time budget instead of one that had already run out. Estimates that used to fail twice in a row now get a real second attempt.

Timed-out requests now retry properly

When a request took too long, the automatic second attempt could fail instantly instead of genuinely trying again. Retries now get their own time budget, and when there is not enough time left for one you see the real error straight away instead of a second, unhelpful one.

Timed-out requests now retry properly

When a request took too long, the automatic second attempt could fail instantly instead of genuinely trying again. Retries now get their own time budget, and when there is not enough time left for one you see the real error straight away instead of a second, unhelpful one.

Timed-out requests now retry properly

When a request took too long, the automatic second attempt could fail instantly instead of genuinely trying again. Retries now get their own time budget, and when there is not enough time left for one you see the real error straight away instead of a second, unhelpful one.

See the numbers behind the recommendation

When your scenario includes figures - a bonus percentage, a monthly premium - the advisor now works out what they come to over a year and shows them with the analysis, in the Overview tab and in the exported PDF. Each one is checked against the figures you entered.

Advanced proposals finish instead of erroring

Generating a proposal in Advanced mode had been failing with an error rather than finishing. It now completes. Advanced takes longer to write than Simple — around two minutes — because the model behind it is doing considerably more work.

Your quote totals only the services you listed

Sometimes the proposal added a line item you never asked for — and priced it into your total. The pricing table now shows exactly the services you entered, at the prices you set, and the total adds up only those.

Your payment terms are used as you wrote them

Ask for a 40/60 split and the proposal now says 40/60 in both the payment schedule and the terms section. It used to quietly fall back to 50/50 in one of them and contradict itself.

Advanced analysis finishes instead of timing out

Advanced mode on a longer contract could run out of time and come back with nothing. It now has the time it needs to finish. If a run ever does come up short you get a clear message instead of a blank error, so you know to try again rather than guess.

Clearer message when no file is attached

Sending an invoice to the tool without attaching the image returned a confusing server error. It now says plainly that an uploaded JPG, PNG or WEBP is what the tool needs. This affects the API, extension and assistant integrations rather than the web page.

Premium now delivers the 1080p MP4 it promised

The Premium render is fixed end to end and produces a true 1080p60 MP4 you can download; allow about five minutes. Premium and Long-form stories were also being generated at the Advanced setting despite the higher price, and now run at the tier you chose.

Verification and share links are now free

An Advanced story was priced at 4 tokens but really took 5, or 6 with a link or topic source, because checking and sharing were billed separately. Both are now included. The footer shows the whole price: Basic 1, Advanced 4, Long-form 7, Premium 14.

Explanations tell you when one did not work

Asking for an explanation of a wrong answer now retries on its own if the first try fails, and if it still cannot be written, that question says so. Explain-all used to leave silent gaps while telling you everything had loaded.

Role suggestions no longer fail in silence

Suggest roles from my resume now retries by itself when a request fails, and tells you if it could not finish. It also says when your resume simply did not point to any particular roles, instead of showing nothing either way.

The consulting interviewer speaks again

One of the video interviewers - the consulting engagement manager - had a voice setting that did not exist, so she never appeared on video and quietly fell back to audio only. She now speaks and appears like the rest.

Every interviewer has their own face

Two of the video interviewers - the consulting engagement manager and the chief executive - were played by the same person in different clothes. The executive interviewer now has a face of their own, so each persona looks as distinct as it sounds.

RPG games are now priced by party size

An RPG dungeon crawl used to cost a flat 6 tokens. It now costs 6-12, based on how many characters you create: a solo run stays at 6, a five-person party pays 12 for the extra portraits it generates. You are charged once, at setup, before anything is generated.

AI failures no longer end in silence

Theme generation and prize optimisation used to stop without a word when the AI call failed. They now retry on their own, and if that still does not work, offer you a different AI to try instead — with the cost shown before you switch.

Sources carry their date, or say they have none

A result shows the year it was published when the page states one, and says "no date shown" when it does not. If a source is older than your field usually likes, we say so before you cite it. Dates are never guessed - a missing one is reported as missing.

See what kind of source you are about to cite

Every search result now says what it is - an official body, a journal, somebody's blog, a shop - so you are not judging it from the web address. In a medical, legal, financial or research document we also point out a source that is not the usual kind for that field.

Spreadsheets open properly in Excel

Your interview CSV opened as unreadable characters in Excel if it contained non-English text. It now opens as written.

Spreadsheets open properly in Excel

A CSV export with non-English page titles opened as unreadable characters in Excel. It now opens as written. Password exports are unchanged.

Spreadsheets open properly in Excel

A CSV export with non-English text opened as unreadable characters in Excel. It now opens as written.

Spreadsheets open properly in Excel

A CSV export with non-English text opened as unreadable characters in Excel. It now opens as written, and naming your process in your own alphabet no longer breaks the download.

Spreadsheets open properly in Excel

A CSV export with non-English text opened as unreadable characters in Excel. It now opens as written.

Spreadsheets open properly in Excel

A CSV export with non-English text opened as unreadable characters in Excel. It now opens as written.

Downloads named in any alphabet

Naming an invoice in Ukrainian, Greek or accented letters used to fail the download outright. The file now saves, keeping your name where your software supports it.

Spreadsheets open properly in Excel

A CSV export with non-English text opened as unreadable characters in Excel. It now opens as written.

Spreadsheets open properly in Excel

A CSV or Excel export with non-English text opened as unreadable characters in Excel. It now opens as written.

The transcript reader is shared with Interview Coach

Both coaches read your speech through the same code now. A phrase the browser corrected after settling on it could be recorded twice here; it is counted once.

Words no longer drop out of your live transcript

Speaking in long unbroken sentences could lose short words from the transcript, and the answer you were scored on. Words are kept once the browser settles on them. Text still changes while you speak — that is the browser deciding what it heard.

Your documents keep every letter you wrote

Ukrainian, Greek and accented European text used to come out as question marks in a download. PDF, Word and PowerPoint now print it as written. Chinese, Japanese, Korean, Arabic, Hebrew and emoji are still not supported.

PDF uploads are read again

Every PDF upload had been coming back as unreadable since 13 August, whatever the file. They are read properly now.

Your estimates keep every letter you wrote

Ukrainian, Greek and accented European text used to come out as question marks in a download. PDF and Word now print it as written.

PDF uploads are read again

Every PDF upload had been coming back as unreadable since 13 August, whatever the file. They are read properly now.

PDFs look slightly different, and are larger

The new lettering is not identical in width to the old, so a long document may run a page longer than it used to, and files are bigger.

Your proposals keep every letter you wrote

Ukrainian, Greek and accented European text used to come out as question marks in a download. PDF and Word now print it as written.

PDF uploads are read again

Every PDF upload had been coming back as unreadable since 13 August, whatever the file. They are read properly now.

Scores before and after 22 August are not comparable

Because the volume measurement changed, a delivery score from an older session was measured on a different scale. Compare sessions from 22 August onwards.

A microphone check that matches what you hear

Speaking normally now moves the meter and confirms your microphone. Before, a quiet room could read full and a real voice could read almost nothing.

Your exports keep every letter you wrote

Ukrainian, Greek and accented European text used to come out as question marks in downloads. PDF, Word and PowerPoint now print it as written.

Skip stops the interviewer talking

Skipping a question now silences the interviewer straight away, including on the last question, where the voice used to carry on.

Pronunciation that plays every time

Some words came back silent after a wait of several seconds. The clip is now checked before it reaches you, and a silent one is replaced rather than played.

Moving on stops the last word playing over the next

Clips from a card you have left no longer play over the one you are on.

The screen no longer answers a question you changed

Change something and go again and you get the new result. Slow replies from a run you have moved on from are ignored instead of overwriting the screen.

Your recording plays the whole interview

The download used to stop after the first answer, however long you practised. It is one continuous recording now, and the last answer you gave is in it.

Typed answers accept the words you would actually use

Typed Recall could mark a correct synonym wrong on one card and right on the next. Every card now carries alternative wordings, so the same answer is treated the same way.

A microphone check that matches what you hear

The level meter measures loudness properly. A quiet room reads low and your voice reads high, which is the opposite of what it did before.

Charts now work on the standard plan

Charts needed the premium model and never came out on the standard one. Turning them on now writes the whole document with the premium model, at the premium price, and the box tells you what it costs before you tick it.

Put a picture behind the page

Set artwork on the cover, on every page after it, or on one section. Choose a full bleed, a band across the top, or a panel down the side. A wash goes over the picture before the words, and you can adjust it until it reads the way you want.

Or have the tool draw the artwork for you

Pick a look and it draws a background sized to the shape you chose. It suggests the look that suits your subject before you pick. Uploading your own stays free; a drawn one costs the same as any other artwork.

Titles no longer print over the banner on a PDF cover

A title page with a banner image could print the title across it in the PDF while the Word file looked right. Both covers now agree, and full-width pictures no longer run into the right margin.

Charts never work out a number you did not give them

A before-and-after shows the movement without printing the difference. A funnel shows the drop-off without a conversion rate. A gauge shows how close you are without a percentage. If your own material did not state a number, no chart of it does either.

Eight more kinds of chart, picked to fit your material

Trend lines, dated timelines, before-and-after, drop-off funnels, a two-by-two, cycles, hierarchies and a gauge against a target, alongside the bars and bubbles already there. The tool picks whichever answers what your reader wants to know. Still one token.

Charts are readable in Word downloads again

Every label, number and key inside a chart came out as small empty boxes in the Word download. Screen, PDF and email were always fine. The charts now carry their own lettering, so they read the same everywhere.

Put each chart, table and picture where you want it

Pick the paragraph any chart, table, box of points or photograph should follow, instead of everything sitting at the end of its section. So a sentence that says "the chart below" has the chart below it, in Word, the PDF and the web page alike.

End a page wherever you choose

One click under any paragraph ends the page there, and the next paragraph starts a fresh sheet. Useful when a section should not be split across a fold, or when a summary deserves a page to itself.

A bubble chart for shares of a whole

A new chart for showing how big things are next to each other. The bubbles sit in a cluster, varying in size and position and never overlapping, with a key naming each one. They are scaled by area, and a zero is drawn as an outline, never a small circle.

A picture at the top of a box of points

Boxes of short titled points can lead with one of your own pictures, above the title, the way a printed newsletter runs a photograph over each column. Choose any picture you already placed in that section. Email versions show the boxes without pictures.

Lay a page out in boxes, not just paragraphs

Short titled points now come out as a row of boxes instead of run-on paragraphs, and a key line can sit in a highlighted band. Both print properly in Word, PDF, the web page and email, and stay readable on a black-and-white printer.

What the preview shows is what the file prints

The box layout is worked out once and used by both the preview and the download, so five points never show three-across on screen and two-plus-three in the Word file. A box is never split across a page in the PDF.

Word files keep their headings when you import them

A Word document records which lines are headings, and that is now read when you bring one in. It makes matching your own document far more accurate. PDFs and plain text are read from the writing itself.

Match a document you already wrote

Upload something you have already written and its headings become this document's headings. Two tokens, quoted before the click. Nothing inside it is quoted or cited; it is used as a shape only. If it has no repeatable shape you are told, not handed an invented outline.

Have it ask you questions before it writes

Tick one box and it reads your brief first, then asks three to five short questions about what it could not work out - an in-house acronym, who the reader is, what the deadline is. Answer what is useful and skip the rest. Blanks stay blank. 1 token, and the demo can use it.

Your answers reach the writing, not only the search

What you tell it in the questions goes to whatever writes the document, on every path - including polish, where nothing is searched at all. In research mode it steers the search as well, so a term only used where you work is looked up the way you meant it.

Newsletter and one-page briefing structures

Two new structures for the two shapes people were writing without one. Every structure now explains what it is for, what you get, and when a different one would suit you better.

Tell it what your topic actually is, and search again

If the results are about the wrong thing, say what you mean in your own words - especially if a term in your brief is an abbreviation used where you work. It searches again with your correction. 2 tokens, quoted before you click.

An image dropped on the file uploader is treated as an image

Putting a photo into the "upload a file" box returned a read error that never mentioned the image control further down. It now goes straight to your images.

Longer documents, and a warning when one comes back short

You can now ask for up to 5,000 words, and the writing budget follows what you asked for instead of staying fixed - which is what could leave a long document missing its body. If the result is well under your target, it says so.

Your attached images now reach the document

Attaching a photo could fail without ever mentioning images, and larger pictures never made it into the download at all. Pictures are now resized in your browser as you attach them, and if a file is still too large we name it.

Say what each image is for

An image can now be a page header, a figure inside a section, or a full-page activity on its own page. Pick the purpose when you attach it and every format honours it.

Online research shows you what it found before it writes

Research mode now searches first and stops. You see every result, tick the ones you want cited, and nothing goes into your document until you do. Anything that does not look like your topic arrives unticked with the reason. Reviewing costs nothing.

Search results are checked against your own material

The search now sees the documents you supplied, not just your brief, so it can tell what your terms mean before it looks anything up. Results that share no vocabulary with your topic are set aside, and you are told which ones and why.

Sharper checking of new word sets

A second AI now works out what each word means before it looks at the translation we generated, rather than judging the translation on its own. That catches a word that reads perfectly but means something slightly different.

Industry Deep Dive, on Advanced

Advanced assessments now include how AI adoption is moving in your industry, what waiting costs, and what leading the change there involves. Where the answer really depends on how your own organisation works, it says so and asks.

Answers we could not hear are not scored

If speech recognition drops out for most of an answer, that answer is left out of your score and named on your report instead of quietly dragging it down. Your other answers are scored as normal.

Pick your languages without a system dropdown

The "I speak" and "I want to learn" pickers are now buttons showing each language in its own script, so the list matches the rest of the page instead of being drawn by your operating system.

Your downloaded report matches your screen

Advanced sections — executive summary, ROI estimates, risk register, dependencies, industry deep dive and benchmarking — were missing from the PDF and partly missing from PowerPoint and Excel. All of them are now in all three.

Tell us when a translation looks wrong

Spotted a card that does not mean what it says? Tap "Translation looks wrong?" on the Learn screen or a revealed flashcard. It leaves your deck and your progress straight away, and we review it. Free, and it does not use any of your credits.

No score when we could not hear you

If speech recognition drops out for most of a recording, we no longer put a number on it. You get the transcript we did capture, an explanation, and the option to record again — and the run is not charged.

Pick the interviewer without a system dropdown

Interviewer role is now a row of buttons, matching the interview length control above it, so the choices are visible at a glance instead of hidden behind an operating-system menu.

Calibration says what it actually needs

Calibrating against your own material is an Advanced mode feature. On a paid plan in Basic mode it said you needed to upgrade, which read as a billing problem. It now tells you to switch modes.

PDFs that used to be rejected now upload

Our PDF reader could not open files saved by some modern tools, and failed without saying so. It has been replaced. A file we genuinely cannot read now says so and suggests what to do, and only a scan is called a scan.

The error overlay covers the whole screen

When something failed, the dark overlay stopped short of the bottom of the window and left a pale strip below it. Fixed everywhere that message appears.

Quick Start keeps what you have written

If you have already filled the form in, a Quick Start template now asks before replacing it. Keep your own answers and it still applies the template settings. On an empty form it applies in one click, as before.

Quick Start keeps what you have written

If you have already filled the form in, a Quick Start template now asks before replacing it. Keep your own text and it still applies the template settings. On an empty form it applies in one click, as before.

You are told when a language will not work

Transcription is done by your browser, and not every browser offers every language. If yours refuses the one you picked, you are now told on screen before or during recording, instead of finding out from a strange transcript.

Quick Start keeps what you have written

If you have already filled the form in, a Quick Start template now asks before replacing it. Keep your own text and it still applies the template settings. On an empty form it applies in one click, as before.

Check your microphone before you record

A mic test on the setup screen shows a live level while you speak and tells you whether you are being heard, too quiet, or not heard at all. It only listens when you press the button.

Upgrade notes name the mode, not the plan

Several messages said you needed to upgrade when what you actually needed was to switch to Advanced mode. They now say which it is, so a paid account is never told to pay again.

You are told when a language will not work

Transcription is done by your browser, and not every browser offers every language. If yours refuses the one you picked, you are now told on screen before or during the interview, instead of finding out from a strange transcript.

Quick Start keeps what you have written

If you have already written something, a Quick Start template now asks before replacing it. Keep your own text and it still applies the template settings. On an empty form it applies in one click, as before.

Quick Start keeps what you have written

If you have already filled the form in, a sample now asks before replacing it. Keep your own answers and it still applies the rest of the sample. On an empty form it applies in one click, as before.

Quick Start keeps what you have written

If you have already written steps of your own, a Quick Start template now asks before replacing them. Keep your steps and it still applies the template settings. On an empty form it applies in one click, as before.

Quick Start keeps what you have written

If you have already filled the form in, a Quick Start template now asks before replacing it. Keep your own text and it still applies the template settings. On an empty form it applies in one click, as before.

Your report fits a phone screen

On a narrow screen the export buttons ran past the edge of the report and the feedback icons were squashed to nothing. Both now wrap and stack properly.

Download MP4 now gives you a file

The button could fail silently and leave you with nothing. It now uses the converter that already worked in Public Speaking, and if conversion is not possible you get your original recording plus a short explanation.

Quick Start keeps what you have written

If you have already filled the form in, a Quick Start template now asks before replacing it. Keep your own text and it still applies the template settings. On an empty form it applies in one click, as before.

Check your microphone before you start

A mic test in the setup panel shows a live level while you speak and tells you whether you are being heard, too quiet, or not heard at all. It only listens when you press the button.

Quick Start keeps what you have written

If you have already filled the form in, a Quick Start template now asks before replacing it. Keep your own text and it still applies the template settings. On an empty form it applies in one click, as before.

Quick Start keeps what you have written

If you have already written something, a Quick Start template now asks before replacing it. Keep your own text and it still applies the template settings. On an empty form it applies in one click, as before.

Quick Start keeps what you have written

If you have already filled the form in, a Quick Start template now asks before replacing it. Keep your own answers and it still applies the template settings. On an empty form it applies in one click, as before.

Quick Start keeps what you have written

If you have already filled the form in, a Quick Start template now asks before replacing it. Keep your own answers and it still applies the template settings. On an empty form it applies in one click, as before.

Results fit a phone screen

The four summary figures and the button beside them overlapped on a narrow screen. The figures now sit in two rows of two, with the button clear beneath them.

Quick Start keeps what you have written

If you have already filled the form in, a Quick Start template now asks before replacing it. Keep your own text and it still applies the template settings. On an empty form it applies in one click, as before.

Camera on means camera on

Turning the camera on during an interview now shows you straight away and records the session, so the body-language review is there when you finish. It starts off, because it records. A blocked camera now says so, instead of looking like one you switched off.

Practise in the language you speak

A language picker covering 46 languages, so your speech is transcribed in the language you actually used. Filler-word counts and speaking-pace guidance only work in English, so in other languages we now hide them rather than show a misleading score.

Practise in the language you speak

Choose from 46 languages before you start, and your answers are transcribed in that language instead of always being heard as English. Where a measurement only works in English, we now leave it out rather than show you a number that means nothing.

A wrong answer moves on when you do

The pause after a wrong answer now shows its countdown and can be skipped with a Next button, instead of feeling like the button had not registered.

Your exports actually download

PDF, Word and CSV reports were being built and then lost before they reached your computer. They now save properly. If one ever does fail, you get told why instead of nothing happening.

The scoring screen shows its progress

Reading a full interview takes about a minute. That wait now counts up and names what it is doing, so it no longer looks like the page has frozen.

Sensible answers are accepted again

Typed answers were being checked against alternatives stored in the wrong language, so correct answers were marked wrong. New word sets store them correctly. Sets you built before today stay strict until they are rebuilt.

Long waits show what they are doing

Building your lexicon and running an assessment both take a while. Both now count the seconds, name the stage they are on and move a progress bar, so a normal wait no longer looks like a crash.

A full brief is no longer called empty

A short brief was being told it was required even though you had filled it in. It now says what it actually needs: at least a sentence about what you want written.

Batch pricing now matches what you are quoted

Applying a batch is charged at the price shown for each change you accept, previewing stays free, and downloading your finished file is free again. Refunds for files that fail can never exceed what you were charged.

Run it as often as your plan allows

The hourly limit was five runs, which stopped subscribers long before their daily tokens ran out. It is now sixty. Your token allowance is the only thing that decides how much you can do in a day.

Run it as often as your plan allows

The hourly limit was five runs, which stopped subscribers long before their daily tokens ran out. It is now sixty. Your token allowance is the only thing that decides how much you can do in a day.

Search as often as you need

The hourly search limit was set far too low, and because of how requests were counted it was shared across everyone rather than being yours alone. It is now sixty searches an hour, per person.

Your answers no longer lose words silently

Speech recognition runs in your browser and drops out on an unstable connection. It now reconnects properly, shows a live "reconnecting" prompt so you can pause, and flags an answer if gaps cost you words — so feedback is never based on a fragment without you knowing.

Your transcript now tells you when it missed something

Speech recognition runs in your browser and drops out on an unstable connection. It now reconnects properly, shows a live "reconnecting" prompt so you can pause, and notes above your score if gaps cost you words — so a low score is never quietly a connection problem.

Body language scoring is part of Advanced

Body language scoring is part of the Advanced plan, matching the plan comparison, and costs no tokens there. Other plans see it clearly locked instead of silently missing.

Body language scoring follows your plan

Body language scoring is part of the Advanced plan. It costs no tokens, so it no longer switches off once your daily allowance is spent.

Interviews start reliably again

Practice interviews could stop with a "rate limit reached" message part-way through a session, even in Basic mode with tokens still available. You can now run many more per hour, and an automatic retry after a dropped connection no longer counts against you.

Body language scoring runs with your webcam

Record with your webcam on Advanced mode and body language scoring - gestures, eye contact, expressions and posture - is now included by default. It appears in the Delivery tab. If it cannot run, the page now tells you why instead of leaving the section out.

Revising a section updates the checks with it

After you revised a section, the source summary and the writing check still described the version you had just rewritten, quoting sentences no longer in the document. Both now follow your edits.

The patient-data check runs on every document

It only ran when you chose the Medical style, which is not the default. It now looks for identifying details whatever kind of document you are writing, and no longer mistakes a heading for a name.

The preview is arranged like the file

Charts and tables appeared in one order on screen and the other in the exported document. The preview now matches, so what you proofread is what you send.

The demo says free, because it is

The free demo was showing a token price for a run that never costs anything. It now says free, and shows what the same setup would cost once you sign in.

Charts are visible in dark mode

Charts were drawn in dark ink on a dark card, so they were present in the preview and impossible to see. They now sit on paper in both themes, exactly as they do in the file.

Artwork you paid for is no longer the first thing dropped

A document can carry eight pictures. When you were over the limit, the generated artwork was cut and your uploads kept. Illustrations now come first, and the button tells you when you are at the ceiling.

Required notices now print in your document

Documents that must carry a legal or professional notice were showing it on screen but leaving it out of the file. Word, PDF, rich text and email now all print it.

Tables you can edit, not pictures of tables

Rate cards, schedules and specifications now come out as real tables: editable in Word, ruled in the PDF, and proper markup in the rich text and email versions.

Charts now reach the file, not just the preview

Figures drew correctly on screen and were missing from every download. They now appear in the Word file, the PDF, the rich text and the email version, with their captions and sources intact.

Change what came back, without starting over

Rewrite one section by describing what should be different, redraw a single figure, or add artwork to a section. Rewrites are free. The redraw and the artwork are quoted before you click and charged only if they work.

Loading a template no longer wipes your draft

If you have already written something, loading a template asks first and keeps your text while still applying the template settings.

Medical documents refuse patient data

Health information is now detected in your brief and in your pasted sources, and blocked by default. The message names what it found and links to our redaction tool. You can acknowledge and continue if you have the authority to.

Your attached images now appear in the file

Images you attach are included in Word, PDF and rich text. Give each one a caption, choose which section it follows, and set full or half width so two sit side by side. The email version cannot carry pictures, so it names each one in its place instead.

Charts and diagrams in your document

Add process flows, comparisons, statistics and bar charts. They are drawn with real type, so the text stays sharp when you print it. Turn them on with one checkbox; you see every figure before you export.

Your logo on the front page

If you have set a logo on your account, it now prints at the top of the front page - centred above a title page, or beside the title if you are not using one.

Two side-by-side images kept their captions apart

In Word, a pair of half-width charts printed both captions on one shared line, with neither under the picture it described. Each caption now sits under its own.

Choose whether the document says it was AI-assisted

Some organisations have to state that a tool helped produce a document. The line is now yours to switch on or off in the tool, and it is off by default.

Legal and financial writing styles

Two more registers, each with the guardrails the profession needs. Legal explains what the law says and will not write as your attorney. Financial states facts about products and figures and will not promise or project a return.

Memos, minutes, SOPs, press releases and case studies

Pick the form and it lays out the right sections and the fields that sit above them - To, From, Date, Re on a memo; attendees on minutes.

Add a title page

A cover page with the title, summary and date in Word, PDF and rich text. Off by default.

It tells you what is missing instead of making it up

A memo with no date comes back with the date blank and the gap named. Nothing required is ever invented to make the page look finished.

The PDF keeps your punctuation

The PDF was flattening curly quotes, dashes and bullets that the Word file kept, so a checklist bulleted in Word and hyphenated in the PDF of the same document. All four downloads now render your text the same way.

Page numbers on anything longer than a page

Multi-page PDFs carry "Page 2 of 7" so a printed document can be put back in order. Word files get a page-number footer.

It spots the notes you meant to keep to yourself

Your material contains notes as well as content. A TODO, an open question or an unverified aside can end up restated in the finished document, and nothing else would catch it. These are now flagged for you to cut before you send.

Steps and checklists come out as real lists

Procedures, checks and enumerations now format as proper numbered or bulleted lists in every download, with the wrapped lines lined up under the text. The numbering is applied for you, so it stays right when an item moves.

A clean copy to send, a review copy to keep

The flags are your working notes, not your reader's. Every download is clean by default. Tick one box for a review copy that marks each flagged paragraph in the margin and stamps it not for distribution.

Every sentence checked against the sources you gave it

Each paragraph is read back against your own material and marked supported, unsupported, contradicted or opinion. A sentence is only marked supported when a source you provided actually carries it.

Word, PDF, rich text - or a ready-to-send email

Take the finished document out as a Word file, a print-ready PDF, rich text, or an email version with both the HTML and plain-text parts. Formatting costs nothing extra - you already paid to write it.

A writing check that flags, never rewrites

Every document gets checked for the tells of machine writing: filler phrases, repetitive rhythm, and invisible characters. You are shown what it found and why. Your words are left alone.

Turn your own notes into a finished document

Paste notes, upload a PDF or Word file, or add a link, and get back a structured newsletter, report or explainer. It adds no facts you did not give it, and writes shorter rather than padding when your material is thin.

Citations you can actually check

Switch on research and every claim is cited. A citation that does not hold up is removed and the sentence is marked as unsourced, so you can see what is backed and what is not instead of guessing.

Skip a word, clearer audio, real progress

You can now skip a flashcard without scoring it. Pronunciation retries automatically if a clip comes back silent, and the button waits instead of doubling up when you tap twice. Building a dictionary shows real progress, and a mis-tapped quiz answer can be undone.

Your work survives leaving the page

Your prompt, settings and last result now stay put if you navigate away and come back. Quick Start Templates also show which one is loaded.

Find a proposal that did not load

If your connection drops mid-generation, the message now points you to your history — the proposal is saved there rather than lost.

Undisclosed pay no longer counts against a match

When a posting does not state a salary, the analysis now says so instead of judging the role against your target range. A pay range the employer never published is missing information, not a mismatch, and it no longer lowers the match score.

The coach now sets you an exercise every time

Each session it picks a room, a listener, and a rule to work under — "use these three terms", "no jargon at all", "under 35 words" — aimed at whichever audience you are weakest with. Plus three new drills, real documents to interpret, a teach-it-back partner, and a phrase bank of the lines you write that work.

Learn your terms, then rewrite and watch the score move

New Learn mode walks you through each term — how your target level uses it, how to say it plainly, and how people get it wrong — before anything is graded. After an assessment you can rewrite and re-score, with a before-and-after for every audience. Plus a browsable lexicon, past results, and PDF/Word export.

Four audiences read your explanation at once

Write or speak something from your work and see how a peer expert, a director, a non-technical stakeholder, and a direct report each react — with the exact spots where you lost them highlighted in your own words. Build a lexicon for your field and practice daily.

Plain-text résumé & cover letter downloads

Alongside the formatted PDF and Word versions, you can now download your tailored résumé and cover letter as plain text — handy for pasting straight into online applications and ATS forms.

Download your revised résumé — formatted or plain

Your AI-revised résumé is now downloadable as a polished PDF or Word document in three templates (Classic, Modern, and ATS-optimized) — or as plain text to drop into your own layout. No more copy-pasting section by section.

Imports and exports now match your plan

Importing an existing plan, follow-up refinements, and exports now follow your plan. Board-ready PDF and Markdown come with Basic; the 4-sheet Excel risk-matrix workbook is a Pro feature.

Prompt chain step testing now works

The prompt chain designer's "Test step" button now actually runs each step and shows its output — it previously returned an error. Exports (PDF, Markdown, JSON) now also match exactly what your plan includes.

Exports now reflect your plan

SQL and CSV exports come with Basic; Excel, branded PDF, and Visio exports are part of Pro. Export options now match exactly what your plan includes.

PDF perspective fix

Fixed the PDF export so your personalized AI Perspective section now appears (it was showing blank).

Accurate export labels

Exported estimates no longer show a misleading fixed overhead/contingency percentage — just the actual dollar amounts.

Post regeneration fixed

Regenerating an individual post now works reliably — it was failing before.

Bias-aware protection across rewrites and coaching

The bias-aware protection now also covers your AI resume rewrite and coaching Q&A, not just the initial analysis. Exports (PDF and CSV) are aligned to your plan.

Full STAR breakdown on every answer

Your scored feedback now shows the full STAR breakdown — Situation, Task, Action, and Result — on every answer, not just the overall rating. Plus reliability and fairness improvements across scoring, the coaching chat, and the AI video interviewer.

Complete results view + smarter analysis inputs

Results now show everything the analysis generates — implementation roadmap, before/after improvements, detailed metrics, and automation ROI plus suggested tools that were previously hidden. Waste types and interview answers now shape the output; exports match your plan.

Fairer model use + accurate export tiers

Advanced model selection now actually applies to the follow-up chat and section expand/revise. Export formats match your plan. Advanced model labeled correctly (Claude Sonnet 4).

Faster follow-up + clear plan limits

The follow-up chat is now faster and respects your model choice. Batch and export limits now match your plan, and extraction is more reliable on tricky documents.

Paste-text + demo fixed, faster follow-up

Pasting contract text now works reliably (and the public demo works again). The follow-up chat is faster, and report downloads match your plan. PDF/image uploads on Pro now use your full page allowance.

Working report exports + faster follow-up

Report downloads (PDF and Excel) now work reliably. The follow-up chat is faster, respects your model choice, and carries the same anti-bias protections as scoring. Advanced mode now reads more of each resume for deeper, more accurate analysis.

Log & financial sanitizer modes

New Mode presets: All PII, Logs & secrets (IPs, MACs, URLs, UUIDs, tokens, keys), and Financials (cards, IBAN, account numbers, SSN). One click focuses the detectors for the job; added URL and UUID detectors.

RAG chunk export

New RAG Chunks tab: split a PDF's text into overlapping, token-bounded chunks and export as JSONL for retrieval pipelines. Free by default; custom size/overlap and heading-aware splitting on Pro.

Data Dictionary Generator is here

Profile any CSV or JSON dataset into a field dictionary in your browser — types, nulls, uniques, and detected semantics. Add AI-written field descriptions on Pro. Export as Markdown or JSON.

Financial Calculators are here

Amortization schedules, depreciation (SL/DDB/150%DB/SYD), and a currency converter — in your browser, nothing uploaded. Full schedules and CSV export on Pro.

Document Converter is here

Convert between TXT, Markdown, HTML, PDF, and Word in your browser — nothing uploaded. Text, paragraphs, and headings are preserved. Single files free; batch on Pro.

CSV Cleaner & Profiler is here

Profile columns (types, nulls, uniques, stats) and clean CSV data — trim, dedupe, drop empties — in your browser, nothing uploaded. Larger datasets, dropping columns, and dedupe-by-key on Pro.

Data-Format Converter is here

Convert, validate, beautify, and minify JSON, YAML, CSV, and XML in your browser — nothing uploaded. Single files free; batch conversion on Pro.

Image Compressor, Converter & Resizer is here

Convert, resize, and compress images in your browser — and strip EXIF/metadata automatically. WebP/JPEG/PNG free; batch and AVIF on Pro.

Clip & GIF Maker is here

Trim a video moment into an optimized GIF or short clip right in your browser — nothing uploaded. GIF/MP4 free; WebM, longer clips, and higher width/fps on Pro.

Audio Extractor & Normalizer is here

Extract audio from video, normalize loudness (EBU R128), and trim silence right in your browser — nothing uploaded. MP3/WAV free; M4A, AAC, FLAC, OGG, batch, and loudness/bitrate control on Pro.

Video Converter & Compressor is here

Convert, downscale, and compress video right in your browser — nothing uploaded. Choose a quality preset or an exact target size; export MP4, WebM, or GIF free, with MKV/MOV, custom sizes, and batch on Pro.

Professional redaction: OCR, names & true PDF

Now reads scanned PDFs and images (OCR), finds names, places and addresses (with an optional AI name-finder), and can output a truly redacted PDF — flattened so the underlying text is removed, not just hidden. Plus deeper scrubbing of Office metadata, comments and links.

Redact whole files, not just text

Upload .xlsx, .docx, .pdf, .csv, .tsv, .json or .txt and get it back redacted in the same format — all in your browser, nothing uploaded. Reversible tokens stay consistent across every cell, page and paragraph. (Pro)

Eyedropper, contrast & more exports

Pick colors from anywhere on screen with the eyedropper, see readable-text (contrast) indicators on each swatch, and export as SCSS, GIMP, a hex list, or an SVG swatch sheet.

More detectors

Now also catches IPv6 addresses, international phone numbers, IBANs, MAC addresses, private-key blocks, and Bearer tokens — plus broader API-key coverage (Stripe and more).

Lock aspect ratio

A new Lock aspect ratio option keeps the crop box at a fixed ratio while you resize, so it stays matched to your AI tool's exact input dimensions.

Accessibility improvements

Keyboard operation and screen-reader support — the upload area works from the keyboard, swatches and controls are properly labeled, and copies are announced.

Accessibility improvements

Screen-reader support — the text areas, detector toggles, and redaction-style buttons are now properly labeled and announce state.

Accessibility improvements

Keyboard operation and screen-reader support — the upload area works from the keyboard and the tabs, inputs, and extracted-text area are properly labeled.

Accessibility improvements

Keyboard operation and screen-reader support throughout — the crop box moves and resizes with the arrow keys, and controls, sliders, and progress now have clear labels.

Structure-preserving text + faster split

Extracted text now keeps line and paragraph breaks (better for AI and RAG), splitting into single-page PDFs is much faster, and the Keep and Delete tabs no longer share the same page field.

Faster, cancellable exports + ZIP

Multi-output and batch exports now download as a single ZIP (browsers were blocking the extra downloads), long clips process noticeably faster, and you can cancel a running export.

Sharper palettes + RGB/HSL copy

Extraction now surfaces small accent colors instead of averaging them away, you can copy in HEX, RGB, or HSL, and HEIC images show a clear convert-first message.

Clear handling of locked & scanned PDFs

Encrypted or password-protected PDFs and scanned image-only PDFs now show a clear message instead of producing empty or corrupted output.

Color-count slider fix

Fixed the Colors slider so it re-extracts the palette every time you change it, not just the first time.

Reliability fixes

Stitch-back no longer fails on videos with WebM or MOV audio, variable-length WebM clips now load correctly, and unsupported codecs show a clear message instead of a silent error.

Catches more ID and account numbers

Now catches Social Security numbers without dashes, fully redacts long account numbers, and warns you when text exceeds the scan limit so nothing goes unchecked.

PII Redactor / Text Sanitizer is here

Scrub emails, phones, SSNs, credit cards, IPs, and API keys from text before you paste it into an AI. All in your browser — nothing is uploaded. Free covers the core detectors; Pro adds reversible tokens, file upload, and custom patterns.

Brand Palette Extractor is here

Extract a color palette from any image and copy or export it as CSS, JSON, or Tailwind. Runs entirely in your browser — nothing is uploaded. Free gives 5 colors; Pro unlocks up to 12, all export formats, batch, and saved palettes.

PDF Prep for AI is here

Extract text and split, delete, or merge PDF pages to get documents AI-ready — all in your browser, nothing uploaded. Free handles single PDFs up to 30 pages; Pro unlocks merge, delete, split-each, per-page text, and unlimited pages.

Pro power tools unlocked

Pro now unlocks the full set: crop several regions or time segments in one pass, batch-process multiple files, output transparent (WebM) masks, and save your favorite crop sizes as presets. Free stays unlimited up to 720p and 15 seconds.

Video Crop & Stitch is here

Crop any region and time range out of a video, then drop your AI-edited version back in — it is rescaled and placed exactly where it came from, so there is no offset seam. Runs entirely in your browser; nothing is uploaded or stored.

Four new ways to practice

Drill the same words from every angle with four new formats: Listening, Typed Recall (with forgiving spelling), Sentence Cloze, and Unscramble. Each feeds your mastery and can be chained in a single session.

Topic picker and reliable audio

No more blank topic box: pick from quick topic chips and situation starters, or describe your own situation in a full sentence. Pronunciation playback is also more reliable and no longer stops working after extended use.

Smarter stories and level guidance

Story mode now reads sentence by sentence with optional pronunciation and translation aids, plus a graded fill-in-the-blank quiz that counts toward your mastery. Added an audio speed control and a plain-language guide to each level, with a nudge when you are ready to level up.

Story mode and custom sessions

Learn words in context with Story Mode — the AI writes a short story using your words, reads it aloud, and lets you tap any word to hear it. Build sessions from multiple activities back-to-back, choose your quiz length, and keep adding to a dictionary.

AI language coach

Pick a language and level (or take a quick placement quiz), choose a topic, and get a personalized vocabulary set with audio. Practice with flashcards, quizzes, matching, and a memory game while spaced repetition resurfaces words until they stick.

Job board search restored

Fixed an issue where searching real job boards (Google Jobs, JSearch, Adzuna) returned a "premium account required" error and bounced you back to the start. Job board search now works for all signed-in accounts. Company search was unaffected.

More accurate categorization for well-known sites

Sites like Gmail, banks, and cloud drives now land in the right folder reliably, and a bookmark's actual name and address take priority over a messy original folder. Each suggestion shows a short reason on hover.

Resolve uncertain bookmarks faster

Bookmarks the AI is unsure about are grouped together with clickable links so you can confirm or remove them quickly. Pro users get bulk actions and a batch interview that asks one question per cluster of similar links.

Pro: control folder depth and how much gets renamed

Pro users can choose how deeply bookmarks are nested into subfolders, and set how aggressively titles are rewritten, from light touch-ups to a full clean rename pass.

Edit the cleaned bookmark tree like a file manager

Rename, delete, and drag-and-drop bookmarks and folders directly in the results. Search to jump to any bookmark, refine the layout by describing changes in plain English, and undo any edit with one click.

See what each browser added, and sort the before view

When you merge files from multiple browsers, a sources panel shows what each one contributed and lets you filter by file. Sort your original list by name, browser, or folder to find things fast.

Drag-and-drop upload and reliable large libraries

Drag bookmark files straight onto the page to upload, and reset everything with one Start Over button. Very large libraries with thousands of bookmarks now process without dropping any entries.

Smarter sorting for cryptic and untitled bookmarks

Bookmarks with vague or missing titles are now recognized more reliably before sorting, so well-known sites land in the right folder. After cleanup you can review and reassign anything the AI was unsure about.

Save your cleaned bookmarks and reload them anywhere

Save a cleaned bookmark tree to your account and pull it back up on any device without re-uploading. Keep multiple saved versions to compare or restore.

Pro: clean up saved passwords and cookies

Pro users can de-duplicate exported password and cookie files and re-export them in common formats. The cleanup is fully automated, your credentials are never read by AI, and downloads require re-authentication.

Rank folders by how often you actually visit sites

Pro users can add their browsing history so your most-visited sites are surfaced first when folders are organized. Only visit-frequency is used, never your individual history.

Pro: merge browsers, audit exports, broken-link checks

Upload bookmark files from several browsers at once and merge them, catch near-duplicate links, flag broken ones, and export a CSV, Excel, or PDF audit of your library.

AI bookmark cleanup: reorganize, rename, de-duplicate

Upload your browser's bookmark export and let AI sort everything into sensible folders, give cryptic links clear names, and remove duplicates. Try it free on a sample, or run it on your own bookmarks. Re-importable HTML and JSON exports included.

Refactor macros across modules safely

Rename a procedure, move it to another module, or replace a deprecated API across your whole project in one atomic step - every call site is updated together, and the change rolls back cleanly if anything fails.

Reformat tables and reorganize sheets

Rename, hide, reorder, duplicate, and color sheet tabs, and convert plain ranges into proper Excel tables. Renaming a sheet automatically updates the formulas and references that point to it - all free.

Fix conditional formatting and dropdowns

Detect and repair stale conditional-formatting rules and broken data-validation dropdowns. Disabling or deleting them is free; AI-assisted edits are available on paid tiers.

Compare any two saved versions

Pick two versions of a workbook and see exactly what changed - cell by cell, plus named ranges, macros, and charts - in a clear side-by-side diff.

Built-in audit trail for compliance

Every change is recorded inside the workbook itself - what was changed, when, and by whom - giving you a compliance-ready history that travels with the file.

See how everything connects

A relationship view maps the dependencies in your workbook so you can see what a fix will affect. Surgeries propose their full downstream impact upfront, letting you opt out of any cascade before applying.

Preview charts and render them as images

Every chart now shows a live preview, AI chart edits display before-and-after side by side, and you can export any chart as a PNG snapshot.

Ask AI to review your whole workbook

Type what you want in a chat-style prompt and AI reviews the entire workbook - including macro and Power Query code - then proposes fixes you can refine, accept all, or apply one at a time.

Unlock, decrypt, and change passwords

Remove forgotten worksheet, workbook, or VBA passwords without knowing them, open password-protected files, and set a new password - all from a dedicated Protections panel.

Automatic detection of broken items

On upload, the editor flags issues for you: broken formulas, missing references, empty connections, and structural problems in macros and Power Query - so you can jump straight to what needs fixing.

Make quick fixes for free, no AI needed

Common mechanical fixes - clearing broken named ranges, stripping dead links, renaming sheets, removing passwords, and direct edits - now run instantly with no AI step and no charge. AI is there when you want it, not required.

See and edit cells in an Excel-like grid

Browse your workbook in a familiar spreadsheet view with real Excel formatting - number formats, colors, fonts, borders, and conditional formatting. Double-click any cell to edit its value or formula and press Enter to commit.

Clean up to 10 workbooks at once

Drag in up to 10 files, describe one fix, and apply it across the whole set. AI tailors a proposal for each file, you review them progressively, accept or reject per file, and download everything as a single ZIP.

Edit Power Query (M) code

View and rewrite the M code behind your Power Query connections. Ask AI to fix or refactor a query and apply the result straight back into the workbook.

Fix every flagged item in one sweep

When several items in a category have the same problem, a single Sweep all button fixes them in one pass - for example, clearing every broken named range, external link, or empty macro module at once.

Rewrite and manage VBA macros

Ask AI to clean up or rewrite a macro and apply the change directly to your workbook. You can also add, duplicate, and delete individual VBA modules without leaving the editor.

Edit broken Excel workbooks with AI

Upload any .xlsx or .xlsm, see every sheet, named range, chart, connection, Power Query and macro in one tree, and let AI suggest fixes you preview as a side-by-side diff before applying. Then download the cleaned file.

More export formats and steadier generation

Added Mermaid and Visio diagram exports and an optional setup script on SQL output. Generation is more reliable, large schemas no longer get cut off, and resuming restores your full schema and model choice.

Anti-bias toggle for match analysis

New optional toggle (off by default) on the search form asks the AI to ignore protected-class attributes when running deep candidate-fit analysis on a job you select. Turn it on or off per search.

Anti-bias scoring toggle

New optional toggle (off by default) asks the AI to ignore protected-class attributes during candidate scoring and interview-question generation. Designed to align with EEOC and equivalent international hiring guidelines. Turn it on or off per screening.

Anti-bias scoring toggle

New optional toggle (off by default) asks the AI to score gap analysis on role-relevant skills and experience, and not to penalize career gaps from caregiving, military service, illness, or disability. Turn it on or off per analysis.

Bias-aware interviewing

A new optional setting directs the AI to evaluate you strictly on role-relevant skills and ignore protected-class attributes, aligned with EEOC and equivalent international guidelines.

Bias-aware internal talent matching

Two optional controls keep AI match recommendations fair: one strips names and school affiliations before the AI sees them, the other tells the AI to ignore protected-class attributes and "cultural fit" reasoning. Both default off and combine, aligned with EEOC guidance.

Richer chart variety with fewer repeats

Added pie, map, scatter, dumbbell, and heatmap layouts, plus a built-in reviewer that swaps in a better-fitting chart and stops any single style from being overused across a story.

More chart styles and full-story MP4 export

Pick from a wider library of chart and stat layouts, with smart guidance that matches each one to your data. Download your finished story as a single MP4 video ready to post or present.

Long-form stories that stay on track

A new Creative tier plans your story chapter by chapter before writing it, so longer pieces hold their structure from intro to outro instead of trailing off. Export now includes animated clips for every scene.

Video-ready visuals, icons, and one-click export

Stories now use polished scene templates with topic-aware color palettes, thousands of icons, and built-in motion, so they look like a finished video instead of a chart dump. Export the whole story as a slide deck or an image bundle.

Turn data into an animated story in minutes

Paste a dataset, drop in an article link, or pick a research topic, and get an animated multi-scene data story you can play in the browser and share with a public link. Paid tiers add a second AI that fact-checks every scene against the source.

Gradient overlays, themes, and styles

Style your QR overlay with solid or gradient colors, ten one-click preset themes, an opacity slider, custom dot shapes and style templates, and a choice of AI background model.

Guaranteed-scan QR overlays

When pure AI art cannot produce a scannable code, overlay a real QR over the art at adjustable opacity, or generate a background image with the QR on top in your chosen color — always scannable.

More reliable AI art

AI art now generates and downloads reliably: previews no longer stall while generating, every result gets an instant scannable check, and downloads always save the file. A Scannability slider lets you trade artistic freedom for guaranteed scanning.

Lifelike video interviewers

Your chosen interviewer now appears as a talking video presenter that asks each question aloud, with audio narration as an automatic fallback so the session never stalls.

Tailor the analysis and turn it into action

Set your role and intent so results match your perspective, see the reasoning behind each effect, and refine any branch. Advanced users get an action playbook with owners and timelines, plus side-by-side comparison of alternative scenarios.

Clearer insights and sharper questions

The overview now highlights key takeaways, dominant domains, and an overall significance rating instead of restating your situation. The clarifying questions focus on your context, and a live counter shows remaining demo runs.

Trace the ripple effects of any decision

Map cascading consequences of an event up to five degrees out across six domains (economic, social, political, environmental, technological, legal). Paid tiers add deeper analysis, multi-model consensus, what-if branching, and PDF, Markdown, JSON, and diagram exports.

Industry interviewers and pay-per-length pricing

Choose from industry-matched interviewer personas (Tech, Finance, Healthcare, Legal, Consulting, Creative) and pay only for the interview length you pick, with the cost shown before you start.

Comprehensive Gap Closure

Quality Review tab, headline metrics bar, zoom/pan, PNG export, VA/NVA tagging, cycle time analysis, value stream analysis, cost estimates, refinement history with rollback, PowerPoint export, What-If Simulation, interview answers tab.

Quality Pipeline + Before/After + History

Added a validation pass that checks flow logic (ordering, branching, endpoints, loops), a Before/After/Both editing selector, history loading for past charts, calibrated 0-100 feasibility and impact scoring, and better text fit inside shapes.

Process charting & Lean Six Sigma analyzer

Build BPMN flowcharts with swim lanes and color-coded steps, run DOWNTIME waste and DMAIC analysis, get automation suggestions with ROI, compare before/after, import processes, and export to CSV, Visio, PDF, and Excel.

Generate a full disaster recovery plan

Answer guided questions (or import an existing plan) and get a complete DR plan: risk matrix, RTO/RPO targets, recovery procedures, incident command, communication plans, per-scenario playbooks, compliance mapping for seven frameworks, and PDF, Excel, and Markdown exports.

Score, compare, and share your prompts

See a before/after quality score across five dimensions, build negative prompts for image and video models, auto-detect placeholders with a fill-in form, compare ChatGPT, Claude, and Gemini variants side by side, and share prompts by link.

Optimize any prompt with proven frameworks

Turn a rough prompt into a strong one across ten use cases, with framework matching (RACE, CO-STAR, chain-of-thought and more), a prompt-chain designer, model-specific variants, A/B alternatives, and PDF, Markdown, and JSON exports.

Reddit support

Reddit joins as a supported platform with community-focused caption guidance, a dedicated post preview, and Reddit-specific length limits.

Faster, more consistent campaigns

Campaigns now generate in focused stages for faster, more consistent results, with automatic retry if generation fails and reliable coverage across every platform you choose.

More accurate pricing

Pricing is now far more reliable: line totals, discounts, tax, the grand total, and payment-schedule amounts are calculated for you, with market-rate research and verification. Any prices you enter yourself are kept as-is.

More accurate estimates with live pricing

Estimates are now far more accurate: all totals, overhead, contingency, and cash-flow milestones are calculated for you, and prices are researched and verified against live 2026 market rates so the math is always right.

Cost tables and what-if scenarios

Cost methodology now shows clear calculation tables instead of dense text, and a new sensitivity analysis shows how your cost and risk shift under what-if scenarios like a longer timeline or reduced scope.

More accurate, verified cost estimates

Cost estimates are now researched against live pricing and verified after generation, with a badge showing confirmed, partial, or flagged status. Costs are computed from per-task hours for consistency.

Feasibility checks and premium insights

Your plan now flags budget shortfalls, unrealistic timelines, and team skill gaps. Advanced mode adds a follow-up chat to refine results plus Premium Insights — competitive benchmarks, governance, scalability, and multi-year cost projections — with PDF export.

Build vs buy cost comparison

Enter your expected usage and see a side-by-side Year 1 cost comparison of building versus buying, including the break-even point, so you can decide with real numbers.

AI build roadmap generator

Describe your project in a structured form (or start from 6 presets) and get a tailored build plan — whether you need AI or not, ready-made vs custom-build paths, cost estimates, and a risk assessment.

Improve an existing schema, not just start fresh

Upload a schema you already have and get a targeted analysis of what is missing: indexes, audit trails, normalization gaps, and naming issues. The tool fills the form from your file and shows exactly what it changed.

Clean edges and correct orientation

Buildings and areas are now cleanly sliced at your selection boundary instead of warping, with an option to exclude buildings the edge would cut. Models also correctly match the map with north at the top.

Broader map coverage

Many more real-world structures now render — monuments, castles, places of worship, fountains, wetlands, and more — and rivers follow their actual banks. Complex landmarks like the Eiffel Tower render correctly.

Import your own STL models

Replace auto-generated buildings with your own higher-quality STL models, or add custom structures. Rotate, scale, recolor, and drop them onto the surface; imports persist and export in STL and 3MF.

Edit mode and map controls

Reshape your selection with move, reshape, and add-vertex tools plus undo/redo, switch between five map layers, and use a Quick Guide, keyboard shortcuts, and 3D viewer controls like camera reset, wireframe, and screenshot.

AI-assisted area selection

Pick your area without fiddly manual drawing: search by address for an AI-suggested boundary, drop a pin with a block or meter range, describe the area in plain words, or snap your shape to nearby streets.

Enhancement Pack

Ability reference panel, template JSON export/import, ambient music, narrative PDF export, per-question item analysis, GM pause/override controls, team dashboard with cast toggle, and content collaboration via template sharing.

RPG Dungeon Crawl & Multi-Device Team Play

Full RPG game engine with 9 character classes, D20 combat, boss fights, AI-generated portraits, and 4 difficulty modes. Multi-device team play with QR join, GM controls, spectator view, and real-time broadcasts.

Initial Launch

AI-generated quizzes with 6 game formats: standard test, flashcards, race game, tug of war, taboo, and RPG dungeon crawl. QR code sharing for live play. Cross-model answer verification in advanced mode. CSV, PDF, and PPTX export.

Free Tier AI & Templates

Basic AI features for free tier, template save/load system, theme fixes, and token usage tracking.

Game Modes

Secret Santa and Dirty Santa game modes with steal mechanics for holiday gift exchanges.

AI Features

AI document auto-fill, data analysis, prize optimizer, and themed background generation.

AI-guided database schema design

Design a database through a guided AI consultation across many platforms (PostgreSQL, SQL Server, MongoDB, Excel, Dataverse, and more), with normalization and indexing advice, ERD diagrams, and export to SQL, Excel, CSV, or PDF.

Animations & Themes

Slot machine and card flip animations, 8 color themes, and AI-powered branding customization.

Initial Launch

Launched Raffle & Giveaway tool with spinning wheel animation, weighted entries, and CSV import.

Truthfulness & Smart Suggestions

Resume truthfulness checks, pre-validation, AI role suggestions, location radius search, increased character limits, and provider improvements.

Speech-to-Text Transcription

Added speech-to-text transcription for long-form input, enabling hands-free script entry.

Professional mode and templates

Professional mode adds labor-split analysis, material tiers, warranties, and code-compliance notes, plus document import, starter templates, categories, favorites, and a company pricing structure.

Import, edit, and white-label

Import an existing proposal to refine, edit and expand any section, apply your own branding, and start from templates — with a saved proposal history.

Richer rendering and reliable map data

Buildings, rivers, lakes, oceans, and railways now render with more detail and accuracy, and map data loads more reliably for large areas. Rotate models interactively in 3D.

Tiered OCR for better accuracy

Tiered OCR processing for improved accuracy.

Resume & Cover Letter Tools

Resume style options, AI cover letter generation, career gap integration, and exportable job URLs.

Framing & UX Polish

Framing system, resizable panes, adjustable filler sensitivity, video persistence, and export fixes.

Increased document limits

Increased document limits: Basic 10 pages / 50K characters, Advanced 30 pages / 200K characters.

Initial Launch

Launched Multi-Provider Job Search Analyzer with match scoring across multiple job boards and career page directories.

Practice interviews with an AI coach