Changelog: Bug fixes & improvements

Scholarly is always improving. Here you can find every bug fix, major feature, and other improvements we've made.

July 28th, 2026

The right Auto model when chat creates for you

  • When AI Chat starts a video lecture, podcast, slides, flashcards, a visual document, or research, it now uses Auto (Free) on Free and Auto (Premium) on paid plans — the same model choice as the creation dialog.

Better recovery from brief connection problems

  • An account left open in an older browser tab no longer crashes while loading after Scholarly's AI model list changes.
  • PDF processing now retries an interrupted page-image upload before restarting through the slower fallback processor.
  • Video and audio downloads now retry interrupted proxy transfers instead of failing after a brief connection drop.

Choose the AI that writes your podcast

  • Podcast creation now includes an AI model picker. Free uses Gemini 2.5 Flash; paid plans default to Gemini 3 Flash Preview and can choose either. The selection writes and reviews the transcript only — voices and audio generation stay unchanged.

Follow podcasts with chapter slides

  • Every new podcast now includes a visual companion by default for everyone. Scholarly generates one clean Editorial slide per chapter alongside the audio, then follows the current chapter automatically as you listen. You can turn it off for an audio-only episode.

July 27th, 2026

Fewer generations lost to a single hiccup

  • Google Drive files that can't be downloaded are now caught when you pick them, instead of failing partway through a generation.
  • PDFs now open reliably in the in-app reader instead of failing while the browser loads the PDF worker.

Clearer, more honest error messages

  • When a document has no lesson content, we no longer suggest saving it as a PDF again — we point you to the pages or link that do have the material.
  • A Drive file whose owner turned off downloading now explains that and how to fix it, instead of asking you to reconnect an account that was working.

A cleaner, context-aware upgrade screen

  • Opening plan options from an Upgrade Plan button no longer shows a limit warning just because the account has used free credits. Limit warnings now stay attached to the action that actually reached the limit.
  • The upgrade dialog no longer shows a bright outline around the entire window when keyboard focus moves into it. Buttons and fields keep their normal visible focus indicators.

Safer defaults and more faithful creations

  • Every AI creation now starts in your saved language, using your browser locale with an English fallback the first time. Changing Language in any creator updates that account preference everywhere.
  • AI Slides now preserve source values and units exactly. When a table is too dense, it leaves out a column instead of inventing summaries, ranges, or units.
  • The welcome tour can now be skipped from its first screen and dismissed with the Escape key.

Honest, broader practice-exam coverage

  • Exam mode now shows how many of your deck's cards actually have practice questions — "N of M cards prepared" — instead of promising questions that were never generated.
  • On paid plans, creating an exam now fills only the missing practice variants needed for that exam's size, so paid exams are no longer restricted to a stale 10-card pool.
  • Free decks show the 10-card preparation limit right where you pick an exam, with the upgrade path next to it.

Video lectures grounded in your source again

  • Prepared video lectures can once again inspect up to two PDF pages the plan specifically flagged for figures, tables, or equations, while general source rereading stays off.
  • The planner can now supply concise on-screen card copy instead of repeating narration; scenes without it retain the existing narration-based fallback.

Bug fixes

  • The cancel-flow pause offer now matches the 3-month pause you can set yourself in billing settings — it was previously capped at 2.
  • If you cancelled because you did not want auto-renewal, the follow-up email no longer offers a discounted subscription; it offers a plain restart instead.
  • Win-back discount codes could expire before paid access ended. Their expiry is now anchored to remain valid for 21 days after access ends.

July 26th, 2026

Ultimate is now Premium

  • Same plan, same price, same features — Scholarly's paid plan is now called Premium everywhere in the product. Nothing changes about your subscription or billing.

Flagship AI credits for Scholarly Laureate

  • Laureate includes 2,000 weekly credits for flagship-model work. Standard and advanced models stay unlimited on both Premium and Laureate, so Premium access is unchanged.
  • Laureate members can see their weekly flagship-credit balance, usage details, and reset time in Settings → Usage.
  • Laureate members can top up flagship credits in a heavy week without changing plans.

Your work stays where you left it

  • A PDF or video now shows the flashcard deck made from it, instead of an empty tab offering to create one you already have.
  • "Jump back in" on the home page now includes your flashcard decks, not just the file you made them from.
  • The free-limits banner no longer appears while your first creation is still generating — it waits until your work is ready.
  • The chat limit now tells you when it resets, so a daily limit no longer reads as a permanent one.
  • Finishing a creation no longer flashes the create form back up with a "Retry" button while it opens your new content.

More faithful study materials

  • PDF and video summaries now follow the language of your source by default instead of forcing English.
  • LaTeX formulas in flashcards and study questions no longer lose commands such as fractions, Greek letters, or text labels during generation.
  • When an AI creation asks to read more PDF pages than one pass allows, it is now told exactly which pages were skipped so it can continue reading them instead of silently missing part of the source.
  • Flashcards now stop when only a link or reference to missing material arrives, rather than inventing a confident-looking deck from general knowledge.
  • AI podcasts now keep a direct, appropriate tone for sensitive human topics instead of forcing cosmic metaphors into every subject.
  • Infographics no longer risk printing internal layout labels such as “TITLE” or “SECTION” onto the finished poster.

Scholarly for Teams, easier to find and start

  • Teams is now a full plan on the pricing page, alongside Free and Premium, with per-seat pricing that follows the monthly/yearly toggle.
  • Team setup is now reachable from the profile menu, billing settings, and plan options — and those entry points open team setup directly instead of sending you to a marketing page.
  • If you sign up or use Scholarly with a work or school email, signup and your home page now mention Teams — one plan and one bill for everyone — instead of the generic pitch.
  • "Talk to sales" on the Teams page is now a form on the page itself, so enquiries reach us without needing a mail client.
  • The Teams page now leads with turning your organization's existing handbooks, recordings, and decks into onboarding and training content.

Fewer generations lost to avoidable failures

  • You can now make a video lecture from a short topic — paste “Krebs cycle” and it generates, instead of being turned away for being too short.
  • If an AI provider declines to read a document on content-safety grounds, we now try a different provider automatically. Legitimate medical, anatomy, and clinical material no longer dead-ends.
  • A PDF with nothing readable in it now says so straight away, instead of showing “still processing” for ten minutes and then failing.
  • Picking a Google Drive file that is over your plan's size limit is now flagged the moment you choose it, with the upgrade path shown — rather than failing after you press Create.
  • A video lecture whose picture ran out before the narration finished is now caught and re-stitched while it is still being made, instead of failing at the very end after several minutes of work.
  • If an AI provider declines to read your source while preparing a podcast, we now try a different provider before giving up — the same recovery study guides already had.

Bug fixes

  • Switching an existing monthly subscription to yearly billing failed every time with a generic "please try again" message. It now goes through.
  • Some new accounts were dropped onto the home page instead of onboarding right after signing up, and could end up marked as onboarded without ever seeing it.
  • If our servers hiccuped in the moment right after you signed in, you could be sent straight back to the login form even though your sign-in had worked.
  • An empty response from the server could sign you out mid-session even though your session was still valid.
  • A failed page navigation could reload the app and silently discard an edit that was still being saved. Reloads are now limited to the stale-app-version case they were meant for.
  • Opening scholarly.so/content directly showed a "Page not found" error instead of your library.
  • Reloading a podcast part-way through no longer re-asks the questions you had already answered while listening.
  • Videos whose file was structurally damaged could still be delivered and then refuse to play in the browser. They are now caught during generation and rebuilt.
  • Video generation could stall for around 20 minutes before recovering, when the generator waited on a step that had already finished.
  • Explore search quietly fell back to non-semantic results when the embedding provider was briefly busy. It now retries first.
  • Requests in flight during a deploy could be cut off instead of being allowed to finish.
  • Opening a dialog now moves keyboard focus into it, and closing one returns focus to whatever opened it. Focus used to stay on the page behind, which left keyboard and screen-reader users outside the dialog that had just taken over the screen.
  • Pressing Enter in the email field on signup now moves to the password field. It previously did nothing at all.
  • Video lecture title and closing cards no longer print an internal authoring note as on-screen text, and on-screen lines are no longer cut off mid-sentence when the card still had room for them.
  • The “NEW” badge on the home page was too light to read comfortably in light mode.
  • Compact library rows painted the same colour as the page behind them; they now read as distinct cards.
  • Help and marketing pages listed slide themes that are no longer offered.
  • The screen where you choose what to make announced “Your content is ready!” before your file had been uploaded. It now names the step instead of claiming your source is already saved.
  • The monthly and yearly options in the upgrade dialog can now be moved between with the arrow keys, the way a set of radio buttons is expected to work.
  • Podcast instructions that name which hosts should appear are now followed, instead of the standard three-host cast being used regardless.
  • Video generation could throw away a nearly-finished lecture and start over after a step briefly misbehaved, even though it had already corrected itself — adding several minutes for no benefit.
  • Audio and video that would not play blamed an expired link and told you to refresh. Nothing expires, and refreshing never helped; the message now says what actually happened.
  • Help lists and navigation no longer perform an unnecessary wait that could contribute to stale or unexpectedly reloaded views.

July 25th, 2026

Smarter defaults and clearer dialogs

  • AI Chat and creation dialogs now offer Auto (Premium) and Auto (Free), so you can let Scholarly use the recommended model for the quality tier you choose.
  • Voice-picker search now uses one clean, consistent surface in both light and dark mode.
  • Plan options now show a yellow limit explanation only when the action that opened them actually reached that limit.
  • Locked Full HD and custom-length video options now explain exactly what Ultimate unlocks before showing plan choices.
  • Welcome tours now introduce what Scholarly can do before showing the option to skip the rest.

Video lectures show your numbers, not placeholders

  • Chart, equation and statistic scenes in an AI Video Lecture now use the real figures from your source. Previously these could fall back to placeholder scaffolding — a generic two-bar chart or the scene's own title typeset as an equation — even while the narration read your actual values.
  • When a scene genuinely has no figure worth showing, it now renders as a clean text callout instead of an empty or invented visual.

Podcasts

  • New podcasts now come with interactive questions. While you listen, a quick check-for-understanding question appears right after the idea it tests finishes — and never pauses the audio.
  • Answer to get an explanation grounded in the episode, revisit any question from the episode's question list, and track your score across attempts. You can switch questions off from the player at any time.

Exams and quiz limits

  • The Exam tab now prepares its own study questions instead of waiting on the Questions tab, so exams open on a deck you have not quizzed yet. If questions cannot be prepared, you get a clear message and a Try again button rather than an endless spinner.
  • Returning to a deck after using your daily practice questions now shows the full limit screen — questions remaining, when it resets, and a way back to your flashcards — instead of a bare notice.

Interface fixes

  • The upgrade dialog now scrolls your selected plan into view on short screens, so the highlighted term always matches the price on the button.
  • On phones, the chat box on Home no longer overlaps its own buttons, so tapping anywhere in the field starts typing.
  • The “How was this?” card after a generation no longer sits on top of the chat send button.
  • Copying a share link from a private item now says up front that it will make the item viewable, instead of only telling you afterwards.
  • Home creation shortcuts are back to the roomier three-card layout, with full AI tool names and the original quick-access order.
  • If Google's permission screen closes without the Google Drive box ticked, Scholarly now tells you which box to tick and offers Try again on the spot — instead of a message that disappeared and left no way to retry.

July 24th, 2026

Teams for training, onboarding, and sales

  • New Teams marketing pages for employee training, employee onboarding, sales enablement, and businesses — plus help docs for getting started, billing, sharing with Your team, and switching from an individual plan.
  • Expanded Teams help covers seats ($45/seat/month or $324/seat/year), weekly credits, and the shared library workflow for L&D and research groups.
  • New ChatGPT Business (formerly Team) comparison and guides for buyers evaluating AI employee training tools and sales enablement from company docs.

Clearer study-tool guides and comparisons

  • Updated PDF Summarizer, flashcard, homework-helper, Quizlet, and Chegg pages with clearer comparisons, examples, and more accessible imagery.

Chat attachments stay with the right message

  • Images and other attachments sent with your first chat message now stay on that message instead of jumping onto the next one.
  • Attaching multiple images in chat now shows a compact equal-height grid (up to four) instead of stacking large previews in a tall list.
  • Fresh web searches in AI Chat now return current sources again instead of failing when the search provider rejects an outdated crawl option.
  • Chat replies no longer fail permanently when one AI model rejects a long conversation — the message now moves to another model instead of leaving the conversation stuck.
  • The chat model picker now highlights four recommended choices, making it faster to choose a capable model without scanning a long shortlist.

Video lectures and research recover instead of failing

  • Video lectures and research runs interrupted mid-generation by a server restart now retry automatically instead of failing and refunding.
  • A video whose final render fails a quality check now regenerates once automatically, and a video ending moments before its narration is repaired in place instead of rejected.
  • Long video lectures no longer stop early against an internal step limit that was smaller than the work they were approved to do.
  • PDF study guides no longer fail, or get cut short partway through, on longer documents.
  • Flashcard sets from dense sources now switch to a working model in minutes instead of retrying until the time limit, so decks arrive far sooner.
  • Video lectures now catch a corrupted or silent render while the video can still be rebuilt, instead of failing at the very end.
  • A video lecture that got stuck retrying the same failed step now gets a clear correction and, if it still cannot move on, is retried automatically instead of quietly using up the whole generation.
  • Videos, podcasts, flashcards, and other AI-created study materials no longer fail just because the model produced slightly more text than requested.

You stay signed in through brief outages

  • A temporary connection or server error while loading your account no longer signs you out and sends you back to the login screen.

More reliable creation from your sources

  • Recording and transcription no longer use an AI creation, leaving your included creation available for the flashcards or quiz you make from the recording.
  • When one source fails while creating Flashcards or a Quiz, successfully uploaded sources stay ready and the failed source can be removed without starting over.
  • Creation dialogs now keep clear limit guidance and an Upgrade action visible when a server-side plan limit is reached.
  • Post-purchase tools and cancellation guidance now keep the record of features you have tried after your account refreshes.
  • Signing up from a tool page now carries that tool into onboarding and opens the matching creation flow, so you can continue without starting over.
  • The welcome tour now introduces upgrading earlier, restoring the step order that helped more new users reach the plan explanation.
  • Slide creation no longer gets stuck repairing harmless text overlap, and Very High flashcard sets once again target at least 200 useful cards when the source supports it.
  • If Essay Writer generation is interrupted after claiming a refund, the system now reconciles it so the AI creation credit is returned.
  • Google Drive searches for large or shared study materials now read every result page and use the file details those filters need.

Clearer billing choices

  • After a successful payment or plan change, your active plan now appears immediately instead of briefly showing the old plan or asking you to retry checkout.
  • The plan chooser can now start with monthly billing selected, making the lower-commitment option easier to continue with while yearly savings stay visible.
  • Monthly members see the yearly-switch reminder near renewal, at most once per billing cycle.
  • If you leave a yearly checkout, the recovery screen now leads with monthly billing instead of immediately repeating the yearly offer.

Smoother dialogs

  • Dialogs no longer pre-select Close, Skip, or another button as soon as they open, removing the unexpected blue outline around those controls.
  • Creation settings now sit in their own labelled section, separate from the design choice, with each option on its own row so the setup is easier to scan.

A much better phone experience

  • Fixed screens where the main button could end up off-screen on a phone: the post-upload "what to create" chooser, the post-purchase welcome steps, the checkout offer, and the practice exam setup now keep their primary action reachable.
  • Long flashcard answers, quiz questions, and the end-of-session summary no longer get cut off with no way to scroll to them, and long multiple-choice options are shown in full instead of being truncated.
  • Plan and competitor comparison tables now stack into readable cards on phones instead of squeezing three columns into a sideways-scrolling strip — and the explanations that were previously hover-only are shown inline.
  • Video, recording, and PDF pages now show the item's name on phones, including when you open a shared link.
  • Hitting your AI creation limit now shows the upgrade options instead of the create buttons silently doing nothing, and the "finish yearly checkout" reminder no longer covers the chat box or recording controls.

July 23rd, 2026

Reliability fixes

  • Long, tool-heavy AI Chat conversations now keep your latest message intact instead of occasionally failing before the answer starts.
  • Large PDF summaries now stay within safe output bounds instead of occasionally failing near the end.
  • Healthy long video lectures are no longer rejected when export creates a harmless sub-second difference between audio and video timestamps.
  • Team workspaces avoid repeated membership lookups during startup, reducing intermittent account-loading failures under heavy load.

A clearer refund policy

  • The refund policy now explains that most requests made within 30 days are approved, while unusually extensive use, fraud, or abuse may require a case-by-case review.

July 22nd, 2026

More room to shape your AI

  • Custom instructions now allow up to 5,000 characters across Chat, Skills, Custom Designs, and every supported create flow.
  • Plus and Ultimate members can now choose Claude Sonnet 5 for AI Video Lectures.

Short-form video, made more reliable

  • Kids & Story is now the landscape twin of Short: a 30-second, full-width illustrated story with crisp cuts, paper-strip captions, and the same logo controls.
  • Autonomous video modes now verify that narration and visuals describe the same chapters before an export can finish.
  • Large and scanned sources are processed in small, resumable batches, so a worker restart does not throw away completed video preprocessing.

Creating and upgrading

  • AI Study Guide is easier to find on Home and after uploads, while Worksheet now lives under More in a balanced action grid.
  • Generated and edited page titles no longer revert to Untitled when you leave the editor.
  • Plan dialogs now show the true reason they opened and support complete Tab and Shift+Tab keyboard navigation.
  • Large textbooks and PDFs now finish processing much faster, so you can start studying sooner.
  • Large PDFs now open incrementally and stay responsive while you scroll or zoom, without downloading and rendering the entire document up front.

July 21st, 2026

New Gemini models

  • Gemini 3.6 Flash and Gemini 3.5 Flash Lite are now available across Chat, Research, Slides, Video Lectures, Infographics, Mind Maps, and Flashcards.
  • If a retired model was selected, Scholarly now chooses the current default for that feature instead of silently pretending it is a different named model.

Clearer plans and recovery

  • Yearly and monthly checkout options once again show the exact charge up front — including $144 today or $30/month — with a direct monthly choice.
  • Returning from checkout now shows the file, action, or limit that led you there and lets you pick up where you left off.
  • Upload and AI limits stay visible on Home, Chat, and page upload surfaces with accurate reset or lifetime wording and a persistent upgrade action.

Your source survives the whole create flow

  • Files are saved before customization or checkout can navigate away, so returning to the library no longer loses an upload that was ready to create from.
  • Practice Exams in Auto language now follow the main language of the source instead of being overridden by a small translated glossary.

July 20th, 2026

Scholarly for Teams now starts at one seat

  • You can create a full Teams workspace for one person and add teammates later; the 29-seat self-serve maximum is unchanged.

A refreshed Help Center

  • A new Memory guide explains how the AI Assistant saves useful facts, how to add memories yourself, and how to mark them outdated or restore them.
  • Every Help Center article was checked against the current product, with outdated model, plan, workflow, and feature details corrected.

Fixes and reliability

  • Chat keeps generating when a phone locks or a connection drops, then reconnects quietly; Stop is now an explicit action instead of an accidental disconnect.
  • Emoji in Word, PowerPoint, and other Office files now render visibly in converted PDFs instead of disappearing.
  • Clearing the email field on login no longer silently clears or desynchronizes the password field.
  • Podcasts, recordings, and Deep Research now show clear Processing or Failed states across Home, libraries, and search.

July 19th, 2026

Mind Maps with real, readable text

  • Mind Maps now render as vector PDFs with exact labels and a native text layer, so their text stays crisp, searchable, selectable, and accessible.

More resilient AI generation

  • Podcasts now save completed chapter audio and retry through temporary voice-provider outages instead of discarding the whole episode.
  • Video Lecture narration is synthesized as one continuous performance whenever possible, removing the repeated speed and tone reset at every chapter.
  • Flashcards can now read browser-printed PDFs from their page images when the selectable text contains only headers, dates, and URLs.

Fixes

  • Email signup or login now offers a real Continue with Google button when the address already belongs to a Google-linked Scholarly account.
  • Branded Short videos now include a crisp Scholarly mark, while paid no-logo exports stay watermark-free.

July 18th, 2026

Smoother AI Chat

  • New answer text fades in smoothly, citations appear as source badges while the answer is still streaming, and the finished response no longer snaps into a different layout.

Flashcard exports you can actually reach

  • Export and Daily Goal now open with a click or tap, making PDF, Print, and Anki export available on both desktop and touch screens.

Podcast reliability

  • Podcasts recover from temporary transcript and voice-generation failures with bounded retries instead of failing the whole episode on the first provider hiccup.
  • Older podcast setups with a missing guest name now safely use “Guest” instead of crashing during transcript creation.

A clearer yearly option

  • Monthly subscribers who cancel because of price can switch to yearly billing inside the cancellation flow, with the full charge shown before confirmation.

July 17th, 2026

Illustrated video lectures

  • Video Lectures have a new Illustrated design: a full-length, hand-drawn explainer where one quirky character sketches every idea on a clean white board.
  • Each scene turns one idea into a simple visual metaphor — a coin ramp, a growing jar, a balance scale — with its handwritten labels drawn right into the art.
  • Both Illustrated and Short now hold each scene as a clean still with crisp cuts, instead of a slow zoom.

Short videos

  • Video Lectures have a new Short design: a ~30-second vertical (9:16) video that distills your source into one tight, illustrated story.
  • The narration appears on screen word-for-word as paper-strip subtitles, synced to the voice — Shorts work with the sound off.
  • Every Short is art-directed in one of four Scholarly styles — cobalt dither, oil painting, papercraft diorama, or sketchbook journal.
  • The video player now shapes itself to vertical videos instead of adding black bars.

July 16th, 2026

Fixes & improvements

  • Video lectures now automatically retry on a different AI model when a provider has a temporary problem, instead of failing your generation.
  • Video lectures that run past their generation time limit now automatically retry once on a different AI model instead of failing and refunding your credit.
  • Flashcard study questions no longer disappear when the AI skips a card — the deck keeps every question it generated and fills in the rest the next time you open study mode.
  • PDF summary errors now show a clear, friendly message instead of technical error text.

July 14th, 2026

New — choose your voice across Scholarly

  • Scholarly now includes 26 official multilingual Grok voices: 21 new voices plus Ara, Eve, Leo, Rex, and Sal, available for AI Video Lectures, AI Podcasts, read-aloud, and live Voice Mode.
  • Choose the exact voice for live Voice Mode and read-aloud in Scholarly AI settings, pin Maya's guest voice when creating a podcast, or choose the narrator for a video. Each option explains its tone.
  • Luna — a gentle, patient voice — is the new default for live Voice Mode and read-aloud. For podcasts and videos, leave voice selection on Auto and Scholarly will choose voices that fit your material.
  • Voice Mode now handles reconnects and interruptions more reliably, without overlapping old audio or losing a turn that still needs to be saved.

Bug Fixes

  • Fixed a crash that could close the page while an AI answer was streaming. Answers that mixed a plain “*” with a math formula — like “5 * 3” next to \times — could take down chat, PDF, and flashcard pages mid-response. They now render correctly.

July 12th, 2026

Bug Fixes

  • Fixed a rare error that could stop the Home page from loading for some workspace members.
  • Keyboard shortcuts (including ⌘/Ctrl + K search) no longer occasionally error on Safari when a password manager or browser autofill fills in a field.

July 11th, 2026

Bug Fixes

  • Copying a share link now works reliably on Safari and iPhone/iPad — tapping “Copy share link” no longer fails with “Couldn't copy the link.”
  • Flashcard generation no longer dead-ends with “try different files” when the AI returns a response our checker can't read — we now automatically retry on a second model instead of failing the whole deck.

July 9th, 2026

Improvements

  • Sidebar layout refreshed: Home, Chat, and Recently Deleted sit at the top; a primary New button and search sit below; Content / Chats tabs filter Recent, Folders, and chats. Recent and Pinned now use the same type-colored rows as Home, and dark-mode sidebar text is easier to read.

July 8th, 2026

Bug Fixes

  • AI Video Lectures that stall while planning now show a clearer message and retry on a stronger model instead of failing as a vague “technical issue.”
  • Flashcard creation no longer dead-ends after a brief AI hiccup with “try different files” — we now retry automatically when the model returns an incomplete or unreadable response.
  • Fixed sign-up occasionally crashing when finishing the last registration step after Google sign-in.

Improvements

  • Grok 4.5 from xAI is now available (replacing Grok 4.3) across chat and create tools. On paid plans it is the default model for AI Video Lectures, AI Slides, Infographics, Flashcards, and Research (free stays on DeepSeek). Generation runs at medium reasoning depth. Previous Grok 4.3 selections continue to work via automatic upgrade.
  • Refined the yearly Ultimate savings banner on Home so the discount and switch-to-yearly offer are easier to scan.
  • Tightened spacing on the Create-with-AI option grid so longer titles like “AI Video Lecture” fit without being cut off.
  • Removed the rotating study tips from creation loading screens — just the status and progress now.
  • Fixed design/style option cards in create modals flashing a light white hover in dark mode.
  • Simplified the create-modal “close while generating” button to a clear Close action (no more “continue studying” wording).
  • The “More credits” model badge (e.g. on Claude Sonnet 5) now only appears for team/enterprise accounts that use weekly credits — not on individual plans.
  • Grok 4.5 in AI Chat always uses thinking (no Instant / off mode) — matching what the model supports, so chats no longer fail when thinking was turned off.
  • Fixed a chat bug where your message could disappear after a failed reply even though it was saved — the original text stays visible for retry.

July 7th, 2026

Bug Fixes

  • AI video lectures now work from photos and scanned documents. Previously, creating a video from an image or an image-only PDF could fail with an unhelpful message telling you to pick a different model — now we automatically read your images and build the video.
  • Fixed AI Flowcharts where the steps could render as solid black boxes that hid their labels — every step now shows as a clean, readable box with its text clearly visible.

Improvements

  • Cleaned up the sidebar. A prominent New button and search sit at the top, with Content and Chats sections for browsing your library, folders, and recent chats.

July 6th, 2026

Bug Fixes

  • Flashcards no longer fail on a brief AI hiccup — if the model returns an empty response, we now retry automatically (and fall back to a backup model) instead of failing your deck and using up a creation.
  • Fixed a rare issue where creating flashcards could fail with an error or get stuck partway — even when your cards had already been generated. Flashcard creation now reliably saves your finished deck.
  • Clearer message when an AI video lecture source has too little text — instead of wrongly suggesting your file is scanned or telling you to “run OCR,” we now simply say the source is too short and to add more detail.

Improvements

  • Tidied mobile sign-up/login forms and fixed a rare Google sign-up crash.

July 5th, 2026

A cleaner AI Meeting Notes page

  • A slimmer, more elegant player — the bulky playback panel is now a single, compact bar, so your transcript, summary, and tasks get more room and the whole page feels calmer.
  • Easier to read — refreshed the layout with more breathing room around your transcript, summary, and action items.

Improvements

  • Chat now quietly rides out brief hiccups from the AI provider — if a response can’t start because the model is momentarily overloaded, we retry automatically instead of showing an error.
  • Signup starts you in the free workspace — the final “How will you use Scholarly?” step now always pre-selects the free “Just me” option, so tapping Continue takes individuals straight into their personal study space. The paid Teams option stays one tap away, just never pre-picked.
  • Share entire folders — folders can now be shared just like any other item. Recipients get a read-only view of everything inside, and folders shared with “Your team” show up in your team’s shared library automatically.

Bug Fixes

  • Clearer messages when flashcards can’t be created — if you paste a short topic or upload a photo with no lesson content, we now explain what to do for that kind of input instead of showing PDF-only advice about “pages” and “Print to PDF.”

July 4th, 2026

Introducing Scholarly for Teams

  • One workspace for your whole group — put a class, department, research lab, school, or company on a single plan. Every paid feature turns on automatically for every member, and everyone works from your shared material.
  • The most capable AI models — teams unlock our top tier of models across every AI feature, with admin controls to choose which model tiers the team can use.
  • Weekly AI credits per member — each person gets their own weekly credits, never a shared pool, so one teammate’s heavy week never eats into anyone else’s.
  • Admin controls in one place — invite by email or link, Member and Admin roles, shared source libraries, and one central bill. See Scholarly for Teams to start a team.

AI Image Playground upgrades

  • High-quality mode — a new Quality selector lets paid plans generate sharper 2K images on models that support it, and shows exactly which models will use it.
  • See which model is generating — while images render (and if one fails), each tile now shows the model producing it, so comparing models side by side is much easier.
  • Removed the Style menu — describe the look you want directly in your prompt instead.
  • Deleting an image now asks for confirmation first.
  • Fixed the remaining-generations counter showing “Infinity free left” on unlimited plans.

AI now reads your PDF pages the way you see them

  • Generations grounded in the actual pages — when building slides, study guides, worksheets, videos, and other content from a PDF, the AI now reads the relevant pages as full page images by default, so diagrams, equations, tables, and handwriting are read exactly as they appear instead of relying on extracted text that can drop or garble them.
  • Infographics and mind maps see your pages too — they now look at your PDF’s actual pages when planning, so visual material like charts and diagrams makes it into the result.

A cleaner, more consistent look across the app

  • Light mode got a big polish pass — hundreds of panels, cards, and labels across chat, recordings, podcasts, flashcards, settings, and more now have proper contrast, so nothing looks washed out or blends into the page.
  • Dark mode fixes everywhere — missing dark styles on hover states, cards, and text were filled in across the app, and equations in the page editor no longer render as white strips in dark mode.

Improvements

  • Print your PDFs and slides — the three-dots menu on PDF and slide viewers now has a Print option that opens your browser’s print dialog with the document ready to go.
  • Pasting a link no longer uses up your free file upload — links you add as sources are treated as sources, not file uploads, so your lifetime upload credit stays available for real files.
  • Added a Skip option to the welcome tour so you can jump straight into the app.

Bug Fixes

  • Quiz and exam questions now test understanding, not diagram trivia — questions built from your material no longer ask you to recall an exact value or timing read off a source’s illustrative graph (like “at what time does the curve peak?”). They now focus on the mechanism and relationships the figure is showing.
  • Fixed upgrading being blocked for some users — clicking an upgrade button could show “An error occurred while creating the checkout session” instead of opening the checkout page. Checkout now opens reliably, and we’ve added extra safeguards so this class of problem can’t block a purchase again.
  • Fixed Home and All Content sometimes reloading more of your library than needed when switching views or filters, so large libraries stay faster and smoother.
  • Fixed the welcome guide overstating how many free AI creations some accounts include — it now always matches your actual allowance.
  • Fixed pacing markers like “[slow]” occasionally appearing in AI video lecture transcripts.
  • Clearer help when a PDF is really just a printed document-viewer page — if you upload a “Print to PDF” of an online document viewer (SharePoint, OneDrive, or Google Docs) that was saved before the document finished loading, the file only contains the page’s web address and a page number, not the actual document. We now spot this instantly and point you to download the original file — across flashcards, quizzes, podcasts, slides, and AI videos — instead of trying for a while and then showing a confusing error.
  • The Print button on the Formula Sheet Generator now works everywhere — it reliably prints just your formula sheet (not the whole page), and if your browser blocks the print window it downloads the sheet for you instead of showing an error. This fixes it on iPad and iOS, where printing used to dead-end.
  • Fixed the billing and connections pages showing a bright white background for dark-mode users.
  • Fixed chart labels in flashcard statistics being nearly unreadable in dark mode.
  • Fixed several menus and drawers feeling sluggish when opening — they now respond instantly.
  • Fixed the “last viewed” time on shared content showing an older visit instead of the most recent one.

July 3rd, 2026

Custom Designs are now free for everyone

  • Save your own look and reuse it — on any plan — create a custom design (your style instructions plus reference images) and apply it when generating Slides, Story Books, Infographics, Mindmaps, and AI Videos. It used to be a premium-only feature; now everyone can build and use them.

Improvements

  • Your most-used dialogs now open instantly — Upload Content, flashcard and quiz creation, the welcome tour, and the plans dialog no longer pause for a moment the first time you open them.
  • Your folder’s name now shows on your phone — opening a folder on mobile now displays its name at the top, so you always know which folder you’re working in.
  • Deleting a custom design now asks you to confirm — removing a saved design used to happen on a single tap, which was easy to trigger by accident. We now show a quick confirmation first, so you won’t lose a saved look and its reference images by mistake.

Bug Fixes

  • You can now remove reference images in Custom Designs on your phone — the remove (×) button on a reference image was only visible on hover, so it couldn’t be tapped on a touchscreen. It’s now always visible on mobile, and the Save / Back buttons stack neatly instead of crowding one line.
  • Saved web pages now work as study material — if you add a saved web page (for example a Google Drive or Google Docs viewer page) as a source, we now pull out its readable text automatically instead of failing. When a page is really just a sign-in or viewer shell with no content, we tell you right away to download the actual file — instead of trying and failing a minute later.

July 2nd, 2026

Voice Input in Chat

  • Speak your message instead of typing it — there’s a new microphone button next to Send in AI Chat. Tap it and the input turns into a live recorder with a waveform that moves as you talk; tap stop and your words are transcribed straight into the message box, ready to edit or send.

Improvements

  • A proper welcome on your phone — new sign-ups on mobile now get a quick, phone-friendly setup: pick what you want to make (flashcards, an AI video lecture, a quiz, or a study guide), then snap a photo of your notes, choose a file, or paste text — and Scholarly starts building it right away. Previously, phones skipped the welcome tour entirely.

Help Center

  • A full refresh of every help guide — all 58 articles were checked against the current product, and 37 were updated so what you read matches what you see: the new model picker, PDF page selection in every create flow, typeset document themes, AI Meeting Notes, Deep Research file deliverables, and current video modes and lengths.
  • Four new guides: Paste Text (turn pasted notes into study material), Studying from YouTube Videos, Deleting & Restoring Content, and Account, Login & Password Recovery.
  • Removed outdated information across help and product pages — including references to a free trial that no longer exists and features that have been retired.

Bug Fixes

  • Scanned PDFs from Google Drive now work for flashcards — a scanned or image-only PDF picked from Google Drive used to fail with a “blank or unreadable” error even when it was full of content. We now read its pages as images (the same way uploaded scans already work), so it generates cards normally.
  • Clearer messages when a source can’t be read — if your Google Drive connection has expired you’re now told to reconnect it (instead of a generic “try again”), and empty files or not-yet-transcribed videos now explain exactly what to fix — across AI video lectures, slides, study guides, infographics, and podcasts. These cases also stop retrying pointlessly and refund your credit right away.
  • YouTube videos no longer fail with a “no transcript” error when building an AI Video Lecture or Podcast — some videos (most often long ones) could report no transcript even though captions were available. Scholarly now recovers the transcript automatically, so generation just works without needing to open the video first.
  • Fixed the billing link in payment and renewal emails — the “billing settings” link in the payment-failed and yearly-renewal emails pointed to a page that didn’t exist. It now takes you straight to your billing settings.

July 1st, 2026

Improvements

  • Paste Text is now a first-class way to add study material — a new Paste Text option on Home and a Paste text tab in the Add content dialog turn your pasted notes, lecture transcripts, or articles into a saved source you can study from, no file needed.
  • Made it clearer that YouTube links are supported: the Upload Content tile now says so, and the Add content dialog highlights YouTube in its link section.
  • Moved Study Guide to the front of the Home create row.
  • Tidied up the Home page create options: the newest AI agents — AI Timeline, AI SOP, AI Flowchart, AI Lesson Plan, and AI Outline — now live under the More menu along with the AI Image Playground, keeping the main row focused while every tool stays one click away.
  • Added a quick Folder shortcut and surfaced AI Chat in the Home create options, and moved Practice Exam into the More menu.
  • Gave the Add content dialog more comfortable side spacing on mobile, so content no longer hugs the screen edges.
  • Folders can now be renamed from the folder page — click the folder’s name (or the pencil icon next to it) in the header to edit it in place, just like renaming a PDF or flashcard deck. You can also generate a title with AI while editing.
  • Redesigned the chat model picker: a focused Recommended Models list up top, with everything else tucked into a dedicated Other Models menu instead of expanding the same list — plus a full-color Gemini logo.

Bug Fixes

  • Fixed AI Video Lectures coming back in English when your source material was in another language — with language set to Auto, the video now matches your source’s language (Spanish notes make a Spanish video). Picking a specific language still works as before.
  • AI Video Lectures now make fuller use of their runtime on dense sources instead of over-compressing, stay closer to your material (no invented facts), and preserve your own script’s wording when you provide one in the instructions.
  • Fixed the Custom Designs editor’s create button staying silently disabled — it now tells you what’s missing, and the design picker shows upfront that custom designs are a Premium feature.
  • Flashcards now work with scanned or photographed PDFs from Google Drive. If a PDF has no selectable text, Scholarly reads its pages as images instead of stopping — so a scanned textbook still turns into a full deck.
  • Fixed chat citations sometimes showing as raw source codes instead of clickable source badges — source links now render correctly on every chat model.
  • Citations now appear as clickable source badges while the answer is still streaming — previously they showed as raw citation text until the response finished.
  • Fixed AI Video Lecture captions appearing as huge paragraph blocks instead of readable subtitle lines.
  • Fixed the Build Flashcards button appearing off-center in the practice exam “What would you like to study?” empty state.
  • Fixed the model picker in chat not showing a hover highlight in dark mode.
  • Fixed Claude Sonnet 5 erroring out when Thinking was turned on in chat — it now thinks and answers normally.
  • A model’s thinking summary now appears above its answer instead of below it.