Tencent's WeVisCut trains a 9B agent to run the whole editing room, using 33x fewer tokens than Opus
tencent · hf · 2026-10-30
Tencent's WeChat Vision team and Xidian University released WeVisCut, a script-to-video editing system.
- Method: a compact 9B policy coordinates video editing tools; editing inversion and verified replay distill executable supervision from 3,750 finished projects into 134,771 training examples.
- Results: 90.5 overall score on 100 held-out tasks, using about 33x fewer generated tokens than Claude Opus 5 per successful output minute.
- The agent recovers from failures and handles the full pipeline from narration to long-form video.
More from Multimodal
- IntReal: 100K Real-World Interleaved Multimodal Samples to Benchmark Unified Models — nanjinguniv · 2026-10-30
- Hand-drawn-style AI short celebrates Korean Seollal with a good-news magpie — narrator_mochi · 2026-10-30
- Kimi launches 3D Make: connects to Blender and CAD tools to build models, scenes, and games — Kimi_Moonshot · 2026-10-30
- Stunning anime cosplay images made with Qwen Image 2.1 — rarugagamer · 2026-10-30
- Animated short made with ComfyUI: ChatGPT poses, Qwen restyle, MiniMax H3 — bseoan · 2026-10-30
- This Pillow Does Not Exist: a one-click AI pillow design generator — andrew_n_carr · 2026-10-30