Update v3.2.8
September 2026 Update, Part 2
Greetings from LETR AI (formerly LETR WORKS)!
Following the v3.2.0 major release on September 7, we shipped eight follow-up updates during September, from v3.2.1 to v3.2.8. The current version as of September 30 is v3.2.8. Here is what changed in each release.
1. v3.2.1 (September 7) - stable long video editing sessions
- The video no longer stalls near the end after more than an hour of editing.
- Expired video source URLs are now refreshed automatically.
- Fixed duplicate workspace notification requests in AI Signal content details, and reduced duplicate requests during loading and initialization.
2. v3.2.2 (September 8) - assetization and webtoon review
- Create a translation project directly from a completed assetization project.
- The webtoon character list is now looked up per workspace rather than per title.
- Webtoon assetization now includes areas without text in the review scope.
- Asset details and previews now open directly in the browser.
- Main screens make fewer requests, and items that fail to load are shown as failed.
3. v3.2.3 (September 14) - reliable saving in webtoon assetization editing
- Webtoon editing and saving: multiple listeners can be assigned and cleared, and first-person settings, effect expression types and image region edits are saved. The save button and save status are tidied up, and when concurrent edits conflict you can compare against the server version and re-apply your draft.
- Resume where you left off: zoom, selection and tab positions are restored, and completed work can be locked and resumed.
- Numbering and issues: sentence and image numbers come from the server. Issues can now be created, resolved, reopened and deleted on image elements, with line breaks preserved.
- Asset details and previews: improved processing status updates, PSD/image previews and linking to existing assetization projects. An assetization project can be created once speech-bubble analysis is ready.
- Languages and RTL: a common language list is used everywhere, and right-to-left display (Arabic, Hebrew and others) is improved for translation, transcription and active subtitles.
- AI Signal: automatic measurement starts after the first competitor is registered, and unmeasured targets are shown in the list.
- Asynchronous processing status and list refresh for image completion requests, a corrected total count in the admin project list, and improvements to subtitle and dubbing editing, Drive upload of deliverables, and the Korean and English manuals.
4. v3.2.4 (September 16) - webtoon assetization editing UX
- Unified the webtoon assetization editing UX and improved search and save guidance.
- Further fixes for playback stalling during long editing sessions and recovery from expired URLs.
5. v3.2.5 (September 20) - video playback recovery and protected subtitle saves
Playback can now be recovered when it stops during video editing, and unsaved subtitle edits are no longer lost during rapid consecutive edits, failed saves or page reloads.
- A failed video URL refresh is shown on screen, with limited automatic retries and a Retry button.
- Playback position, paused state, speed and loop range are preserved when the video source is replaced.
- Edits made while a save is in progress are saved afterwards, and moving between segments no longer waits for the save response.
- Failed saves, timeouts and empty responses are no longer treated as success; the editor shows pending / failed / saved states.
- Unsaved subtitles are kept in the current tab, and after a reload you can recover and save unsaved changes. Refetching from the server no longer overwrites newer unsaved edits.
- Fixed an error with video URLs containing Korean characters or spaces.
6. v3.2.6 (September 29) - shared assets, credits and contracts
New features
- Download the Enterprise contract in Korean or English from the pricing page.
- A negative credit balance is now shown, with guidance on topping up.
- Select shared materials in Assets and see where they are used. The translation quality button has been removed.
- The SRT full-replacement review screen is now connected.
- Existing TB and assetization translation workflows are now connected.
Bug fixes
- The asset list did not refresh after a TM/TB upload.
- Search scope is preserved, and TM/TB upload failures are handled better.
- Asset search can be reset to the original text, with UX improvements to search, upload and detail views.
- Notifications (toasts) now appear at the top center so they do not block the waveform.
- Cue timing is preserved when editing subtitle text or status.
- Fixed the previous subtitle being restored after splitting a segment.
7. v3.2.7 (September 30) - unified speaker dictionary and webtoon editor performance
- The speaker management screen is now part of the LETR AI character dictionary. Search the workspace’s characters and manage their names in one place.
- The video editor suggests confirmed names from the character dictionary and speaker names from existing subtitles as autocomplete.
- Removed repeated speaker dictionary lookups per glossary in project settings.
- Webtoon editor performance: opening a long PSD episode no longer loads every full-size image at once and freezes the editor. Only the visible range and its surroundings are loaded as WebP tiles, and the page list uses small thumbnails. Original resolution and speech-bubble and edit coordinates are unchanged.
- Deliverable status in the project list: project lists and cards now distinguish sentence progress from deliverable generation (generating, failed, complete), and a project whose generation failed can be re-completed from the list menu.
8. v3.2.8 (September 30) - editor loading states and display performance
- The document, video, video asset and webtoon asset editors now show initial loading, error and retry states.
- The video editor shows when the source is being loaded or recovered, and the webtoon editor shows the loading state of visible images.
- Fixed only some video segments being shown when scrolling after a search.
- Compressed webtoon display tiles are loaded only around the visible area, reducing initial image requests and rendering load.
We remain committed to continuous improvement in performance and convenience. See you in the next update.
Thank you.
[Previous updates]