Hello everyone! Following the August 17 online features patch, this update focuses on expanding paper discovery, guided paper introductions, and Q&A. It also improves save management, the Workshop, and loading performance while addressing recently reported issues with the vision model, blackboard images, and voice playback.
Paper Discovery and Daily Recommendations
Added OpenAlex as a comprehensive paper search source, with Unpaywall and Europe PMC integrated to help locate legally available open-access full text.
Daily recommendations are now based on recent popular papers from OpenAlex. You can select a preset research field or enter your own interests in Settings.
Removed the limit on local paper image extraction, allowing long documents to be scanned in full.
Guided Paper Introductions and Story Generation
Guided introductions now explain a paper's background, key methods, experimental results, and main conclusions chapter by chapter, turning difficult papers into a more approachable interactive story.
Improved chapter division, heading-tree recovery, and structure recognition after standard parsing. Stories now follow the paper's table of contents more accurately, with fewer passages assigned to the wrong chapter.
Strengthened the connection between generated stories and the source paper. Characters stay focused on the current chapter and can use relevant data charts and architecture diagrams when appropriate.
Improved generation concurrency, fallback behavior, and loading progress to reduce stalls, missing chapters, and incomplete content.
Paper Q&A
Added a continuous paper Q&A space where you can ask follow-up questions about the entire paper or the current section and review previous conversations at any time.
Answers use the paper text, chapter structure, and recognized relevant images to provide more focused explanations.
Voice playback can be enabled on demand. If enabled while reading, it takes effect from the next answer without interrupting the current one.
Improved validation against the source paper to reduce off-topic or irrelevant answers.
Paper Images and Blackboard
The vision model identifies photos, data charts, architecture diagrams, and other image types, and understands their main content so characters can use paper illustrations correctly during explanations.
Story generation selects relevant images based on the current chapter and explanation.
Blackboard images now provide hover feedback and can be clicked for a full-screen view. Click again to return to the story and inspect chart details more clearly.
Save Management and Loading
Save management now loads lightweight summaries first, moves Steam Cloud reconciliation to the background, and caches the save list.
Reduced repeated file reads and page parsing, significantly shortening the initial wait when opening save management and background selection.
Save management, background selection, the Workshop, and paper-loading screens now use animated skeleton placeholders instead of loading spinners.
Improved migration and recovery reliability for older saves, character assets, and local images.
Voice and Music
Custom voice generation now supports the GPT-SoVITS v2 API and can use the current character's reference audio and reference text.
Music can now be imported by selecting multiple MP3 files at once.
Previewing music temporarily pauses background music and character voices, then restores them afterward to prevent overlapping audio.
Workshop, Characters, and Release Compliance
Improved browsing, publishing, and subscription flows for Workshop characters, backgrounds, sticker packs, and language packs.
The sticker selector now supports mouse-wheel paging, and backgrounds support video and local media files.
Interface and Stability
Improved the liquid-glass interface, title layout, paper table of contents, original-text reader, and loading feedback across the app.
Replaced the in-game return button with a clearer home icon.
Fixed an issue that made initial voice and vision-model requests more likely to time out.
Improved recovery for dynamic modules, media playback, character sprites, and interrupted generation.
Thank you for your continued feedback. If you still encounter any issues after updating, please let us know through the Steam Community.