FireCRT Releases CoinVE-Edit

·2026.08.20 18:41

Key point

FireCRT has released CoinVE-Edit, a multi-instruction video editing model, on Hugging Face.

Details

FireCRT has released the CoinVE-Edit model on Hugging Face. Based on Wan2.1-T2V-14B and Qwen3-VL-8B, this model is characterized by its ability to process 2 to 5 editing instructions simultaneously in a single forward pass.

Trained on the CoinVE-200K dataset, the model controls each edit to apply only to designated spatial regions through instruction-specific mask injection. This enables complex video editing tasks such as Replace, Add, Remove, and Background Change while maintaining overall video consistency.

huggingface open-source video-editing instruction-following qwen3-vl wan2.1

Read originalShare on Threads

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.