Hello everyone! Four days have passed since the August 24 update. This update focuses on improving document parsing, the story interface, and paper Q&A voice features, while also adding loading screen sound effects, character portrait positioning, Workshop and community entry points, and more.
New Story UI
Added a new “full-width visual novel” story UI. The original card-style dialogue box can still be switched back in Settings.
After selecting an option, the game now first shows whether the selected option is correct or incorrect before continuing to the next part of the story.
Improved the story history, results screen, and various in-game control layouts.
Document Parsing & Story Generation
Enhanced recognition for scanned documents, image-based PDFs, and documents with many charts and figures, allowing data charts, architecture diagrams, and experimental results from papers to be more reliably used in subsequent explanations.
Improved retry handling for document image extraction and multimodal requests, reducing parsing interruptions, missing images, and occasional generation failures.
Streamlined the first-time upload guide to make starting story generation more straightforward.
Paper Q&A & Voice
Fixed Japanese voice playback.
Improved batch voice generation and usage handling for paper Q&A, reducing wait times and duplicate requests during consecutive replies.
Fixed an issue where custom Latin-script languages might not take effect correctly when used as the output language.
Characters, Loading & Sound Effects
Fixed an issue where saved adjustments to custom character portraits—such as position, scale, and mirroring—were not applied in the story.
Added a mute button to the loading minigame, and synced it with the mute state in the story.
Added click sound effects for web buttons.
Workshop & Community
Improved Workshop item cards, tool entry points, and GPT-SoVITS-related operations, making browsing and managing content more compact on desktop.
Stability Improvements
Strengthened strategy reuse and fallback handling for hosted parsing, custom AI, and multimodal requests.
Fixed an issue where some distribution resources were not packaged correctly, improving loading stability for document images, character assets, and story content on desktop.
Thank you all for your continuous feedback over the past few days. If you still encounter issues after this update, please let us know through the Steam Community, Discord, QQ group, or other community channels.