Official implementation of InstructX: Towards Unified Visual Editing with MLLM Guidance
Chong Mou, Qichao Sun, Yanze Wu†‡, Pengze Zhang, Xinghui Li, Fulong Ye, Songtao Zhao‡, Qian He
† Corresponding author, ‡ Project lead; Intelligent Creation Lab, Bytedance
🌐 Project Page | 📜 Arxiv | 🤗 Benchmark |
10.9.mp4
- 2025.10.10: Release VIE-Bench.
- 2025.10.10: Release InstructX tech report.
We introduce VIE-Bench (Video Instruction-Based Editing Benchmark), which comprises 140 high-quality instances across eight editing categories.
| Edit Task | Sub Edit Task | Number |
|---|---|---|
| Total | 140 | |
| Local Edit | Object Swap | 25 |
| Color Change | 10 | |
| Add | 30 | |
| Remove | 30 | |
| Global Edit | Style Change | 10 |
| Tone / Weather Change | 5 | |
| Hybrid Edit | - | 10 |
| Reference Base Edit | Reference Base Swap | 10 |
| Reference Base Add | 10 |
