Qwen-Video-Edit repurposes an image editing DiT to edit videos, no video-pretrained backbone needed
qixing_huang · x · 2026-08-18
Qwen-Video-Edit introduces instruction-based video editing by repurposing the pretrained image editing model Qwen-Image-Edit's DiT to operate directly on Wan 2.1 video-VAE latents — no video-pretrained transformer required.
- Method: two tiny projections bridge Wan 2.1's latent space into the DiT's token space, warm-started from the DiT's own input/output layers so a static video embeds exactly like an image; latent frames are laid out as tiles of one big virtual image with grid RoPE, matching the image model's pretraining. Prompts come via Qwen2.5-VL.
- Training: fine-tuned with LoRA or full parameters on Ditto-1M (source, edited, instruction) triplets, then refined by a few steps of Wan 2.2 denoising-enhancement.
- Capabilities: long-video editing with a different instruction per 45-frame segment, and native aspect-preserving portrait video support.
Code and the Hugging Face model are released.
More from Multimodal
- MegaParts: Scaling Part-Aware 3D Generation to 300 Parts via Efficient Modeling — Shanghai-AI-Laboratory · 2026-08-18
- Creative Demo Combines Faces and Timelines — jh3yy · 2026-08-18
- AI Short Film 'DO NOT WAKE HER' Showcases Visual Effects — RealisticValuable484 · 2026-08-18
- MiniMax Character Swap Test: Person-to-Person and Animal-to-Animal Work — Alex-edits123 · 2026-08-18
- Minimax H3 Generates a 'Deleted Scene' From the Doomsday Trailer — beatlepol · 2026-08-18
- Stabilizing renders with depth maps and object tracking — alecubudulecu · 2026-08-18