Innovate Futures @ Benji
patreon
Workflow Update - Wan Vace 5in1 (Ver. 20250522)
🕑 Added 2025-05-21 21:44:25 +0000 UTCPrevious version : https://www.patreon.com/posts/comfyui-workflow-129211762
Current version update:
RemoveBG and White background node change
RenoveBG node new version only have Alpha background, that cause image cannot process further in image resize or pass into Vace Encode.
Now Fill background added.
Mask Edit Group update:
Connect to shared Model loader group, using flowmatch_cuasvid sampling method, it enable to use CausVid Lora + Vace with masking, and quality remain 14B level.
81 frames mask edit, I got 2:10 minutes with 8 steps.
Video Upscaler simplify in 1 group:
All 5 Vace feature groups can use this, select the GetNode from video output, before you Run.
Version 20250522 attached below, have fun.
Comments
Faraz Aslam
Got to say, this workflow is amazing!!
Benjamin Law
Hey Jonathan! For the face detail, I suggest to follow how the official demo do in Vace as well as Phantom. Character image should be a close up portrait. I use something like this , I can get a face just like the reference image as I tested. If I use a medium shot or full body shot, the AI are going to focus on the outfit and the more its far off distance, the face are going to get less attention. Thanks
Jonathan Oh
Thank you always for your workflow. I really love your works. By the way, for this how do you control face detail? I tried several images but the result of face is very poor.
