---
title: VitronEdit
emoji: 🪄
colorFrom: purple
colorTo: indigo
sdk: gradio
sdk_version: 6.30.0
app_file: app.py
short_description: Instruction-based image & video editing (UniVEdit)
python_version: "3.12"
startup_duration_timeout: 1h
license: other
license_name: tencent-hy-community
license_link: License.txt
models:
  - SanRiiiii/UniVEdit
  - tencent/HY-OmniWeaving
---

# VitronEdit / UniVEdit

Gradio demo of [VitronEdit: Universal Visual Editing across Images, Videos, and 3D Scenes](https://huggingface.co/papers/2610.27049)
([code](https://github.com/Vitron-LLM/VitronEdit-Code), [weights](https://huggingface.co/SanRiiiii/UniVEdit)).

It supports instruction-based image editing (`i2i`), reference-guided image editing (`ti_i2i`), and
video editing with or without a reference image (`v2v` / `tiv2v`). It runs the authors' inference code
(`hyvideo/`, `univedit/`, vendored from the official repo) on top of
[tencent/HY-OmniWeaving](https://huggingface.co/tencent/HY-OmniWeaving).
The 3D scene editing mode (`3d23d` + 3D Gaussian export) is not included in this demo.

The code and weights are derived from Tencent HY-OmniWeaving and distributed under the Tencent HY Community
License Agreement (see `License.txt` and `Notice.txt`).

The example images and videos come from the CC0 pools
[linoyts/repo-to-space-example-inputs](https://huggingface.co/datasets/linoyts/repo-to-space-example-inputs) and
[linoyts/repo-to-space-example-videos](https://huggingface.co/datasets/linoyts/repo-to-space-example-videos).