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
Fix the unnatural transactions between each clip. No more weird pause.
FAQ
Comments (13)
can you make it for 2 images constant face to video
First of all, I want to congratulate you on your work. I would like to ask you one thing: is there any way to stop it from changing the woman's face? It changes the woman's face to a standard one that has nothing to do with the original image. Thank you in advance.
Thank you for the AIO Mega WF. But there is a big issue in this WF. We can't choose witch Lora we want to use for each scene.
just add a lora loader between the model and the next step.
@jwcoop3r631 would we use a high or low lora?
@ctcf6 Low works for both, high only works for high. Play with strength.
How to use for t2v?
I need to know also. Every extended generation workflow I've seen is I2V. Starting to wonder if it's even possible with T2V :\
@SLAPY7 @OwnPear8996980 Try this T2V Rapid WF. https://huggingface.co/Phr00t/WAN2.2-14B-Rapid-AllInOne/discussions/122
Can this model perform role replacement?character replacement