贞贞的无限画布v2.0.9

v2.0.9 · Veo Omni and 3D action director release

Highlights

• Upgraded the 3D panorama node into a fuller action-director workflow: multi-actor colored mannequins, 100+ pose presets, direct joint editing, occlusion guides, keyframe timelines, sequence-frame material sets, fullscreen director controls, and natural-language action planning.
• Added a Sora2 Zhenzhen API channel while keeping the existing Sora2 FAL channel intact.
• Split Sora2 into its own category API key setting, so Sora2 no longer shares the independent Veo key.
• Renamed the video model category from Veo 3.1 to Veo and made veo-omni-10s the default Veo model while keeping legacy Veo options available.
• Implemented the Veo Omni protocol through /v1/videos multipart submission with upstream omni_flash-10s, input_reference, and /v1/videos/{task_id} polling.
• Raised upload handling to 20 MB and return structured 413 JSON on oversize files instead of surfacing misleading HTTP 500 errors.
• Hardened canvas persistence with BOM/NUL cleanup, canvas-file fallback scanning, and atomic writes to reduce data loss after crashes.
• Updated the in-app model usage notice for the current image, video, audio, and LLM model groups.

Electron Update

• Built with the standard npm run dist:release workflow.
• Release assets include the NSIS installer, blockmap, and latest.yml for GitHub auto-update.
• Updater installation remains foreground/visible instead of silent.

Verification

• node --test tests\*.test.ts
• npm run type-check
• npm run build
• npm run public:check
• npm run dist:release
• npm run release:verify