A workflow uses the latest Wan2.2 All-In-One-Mage model and Next Scene Lora and ComfyUI subgraph to create an unlimited video generation process with low VRAM. Perfect smooth transaction between different scenes. With consistency improvements in ID and Color. No more weird pause and direction change.
Please check the video for detailed instructions before you start:
If you think this is helpful, please like the video and subscribe for more.
Description
Using the Next Scene lora with the AIO model to better control the consistency in ID and Color.
FAQ
Comments (5)
First version crashes my confyui install, but 2.0 worked great. Imho this is almost a breakthrough, combined with "next scene" lora to create consistent images, it can generate a nearly perfect video with unnoticable transitions and barely any degradation for almost 30 secs. Mindbl0wing.
I really like this concept! I can't seem to get I2I (storyboard gen) to work on my 3080Ti (12GB Vram). I can generate maybe 2 or 3 storyboard scenes at a time before OOM. Also had to bump steps down to 4 which decreased quality i think.
感谢您的教程。希望除了真人以外，也能出一期动漫角色的工作流，再次谢谢您！
I'd like to use this workflow, but these 3 Nodes are missing:
ResolutionMaster
LayerUtility: ImageScaleByAspectRatio V2
WanVideoVACEStartToEndFramein subgraph 'Initial(Do not copy this one)'
And i can't install them, any guesses for an solution?
WanVideoVACEStartToEndFramein: this one is in node pack wan video wrapper by kj