vace

ComfyUI workflows to create smooth transitions between video clips using Wan VACE. Works with video from any model or other source-LTX-2, drone footage, stock video, personal recordings, etc.

ComfyUI nodes designed to help make common video editing tasks with video generation models less complicated. Smooth transitions, extensions, outpainting. Primarily designed for Wan VACE, with some LTX-2 support.

Unified wan2gp installer for NVIDIA, AMD and Apple Mac, with Python 311, Pytorch 2.10 and NVFP4 Support for RTX50. Super Optimized Gradio UI for AI video creation (6GB+ VRAM). Wan 2.1/2.2, Qwen, Hunyuan Video, LTX Video, Flux and OVI and more. (AMD Windows supported by 7900(XT), 7800(XT), 7600(XT), Phoenix, 9070(XT) and Strix Halo)

Evaluation metrics for Multi-Object Tracking (MOT).

Reflection-aware object removal in 3D Gaussian Splatting via render-edit-refit: delete an object together with its shadow, mirror reflection and colour bleed

Identity-preserving local inpainting for long videos: anchor-frame identity LoRA, windowed multi-seed VACE generation, and Viterbi global chain selection for flicker-free occlusion repair.

Perform your shot instead of describing it — turn a rough reference take into pose, depth, layout, focus & cinematography control signals for controllable AI video (ComfyUI Wan Fun-Control / VACE / LTX).