CivArchive
    Wan2.2_Fun_VACE_Control net_ Video Character Replacement _ First Frame to Last Frame - v1.0
    NSFW

    You can click on the link below to try it out directly. If the effect is good, you can deploy it locally

    https://www.runninghub.ai/post/1967682043029184513/?inviteCode=1cqzbf7a

    Fan benefits,register to get 1000 points,daily login 100 points,play 4090!Experience the super power of 48G.

    Hello everyone, this is the latest Wan2.2 VACE model, the workflow of KJ version, which includes functions such as predicting the beginning and end images, controlling video weights, and replacing characters in videos. Before using it, you need to download the KJ quantified VACE model. I will list the specific model list and usage methods one by one in the following text.

    Models that need to be downloaded

    1.Wan2.2Fun-vace

    https://hf-mirror.com/Kijai/WanVideo_comfy/tree/main/Fun/VACE

    Place folder: models \ diffusionmodels

    2.Lightning_T2V-V1.1-A14B-4Step

    https://huggingface.co/Kijai/WanVideo_comfy/tree/main/Wan22-Lightning

    3.Wan2.2_T2V_fp8 HIGH and LOW

    https://hf-mirror.com/Kijai/WanVideo_comfy_fp8_scaled/tree/main/T2V

    4. Ensure that sageattn can be used normally

    First to last frames

    Upload first and last frame images

    controlnet

    Upload character images and reference action videos.

    Role replaced_Mask retrieval

    Upload character images and videos that need to be replaced, set the video frame limit, run once first, and place the green circle on the character and the red circle on the background in Points Editor to adjust the video mask. Confirm everything and click run again, complete.

    Description

    FAQ

    Comments (4)

    lorobeb735666Sep 18, 2025
    CivitAI

    Do you really think people will get back to your FP8 now that GGUF is out?

    AIGC_Singularity
    Author
    Sep 19, 2025

    Yes, after all, it is quantized. If you want to replace gguf, just modify the model output plugin and model to gguf directly.

    bluehandSep 23, 2025
    CivitAI

    whats the GPU requirement? I tried Wan vace before if try to generate longer than 4sec it screw up. and pose transfer change my ref image also.

    AIGC_Singularity
    Author
    Sep 23, 2025

    Require a GPU of 12GB that can generate videos with 540 * 960 pixels. Wan2.2 not only depends on GPU, but also on RAM. If you cannot run it locally, please increase the blocks to swap of WanVideo Block Swap to 40

    Workflows
    Wan Video 2.2 T2V-A14B

    Details

    Downloads
    886
    Platform
    CivitAI
    Platform Status
    Available
    Created
    9/15/2025
    Updated
    4/28/2026
    Deleted
    -