Tencent publishes WeVisCut 9B tool-using agent for long-form video editing
Tencent published the WeVisCut 9B model card on Hugging Face, describing a tool-using agent that turns knowledge articles or narration scripts into multi-minute edited videos. Based on Qwen3.5-9B and trained on executable editing trajectories, the model reads the current editing state and picks one of 12 tools for planning, narration, asset retrieval, composition, inspection, or repair, and it must be paired with the WeVisCut runtime that executes its tool calls. Deployment requires Python 3.11+, an NVIDIA GPU with roughly 24 GB of memory for bf16 inference, and a setup script that installs Node.js, FFmpeg and the renderer, while the code is under Apache License 2.0 and the weights carry separate terms.
The release packages long-horizon video editing as an agentic, tool-calling workflow rather than a single text-to-video model, giving practitioners a runnable reference for multi-minute video production, though the stated capability and quality come from the project itself and were not independently measured in the supplied report.
tencent/WeVisCut Any-to-Any • 9B • Updated 20 days ago • 16.6k • 31
license: other library_name: transformers base_model: Qwen/Qwen3.5-9B pipeline_tag: any-to-any tags:
- video-editing
- agent
- tool-use
- long-horizon
- reinforcement-learning
- qwen3.5
WeVisCut: A Long-Horizon Tool Agent for Edited Videos
WeVisCut is a 9B tool-using agent that turns knowledge articles or narration scripts into multi-minute edited videos. At each step, it reads the current editing state and selects one of 12 tools for planning, narration, asset retrieval, composition, inspection, or repair. It is not a standalone text-to-video model; use it with the WeVisCut runtime, which executes the model's tool calls.
#Tencent#WeVisCut 9B#Qwen3.5-9B#Hugging Face#WeVisCut runtime