(For SOTAI Squad) Mocha vs Wan2.2 Animate: Best Workflow for PERFECT Character Replacement in ComfyUI
Added 2025-11-03 16:35:18 +0000 UTCLink video: https://youtu.be/brRQCVPdyio
A new video character replacement solution just dropped — delivering results even better than KlingAI, and it’s completely FREE!
In this video, I’ll show you how to build a ComfyUI Mocha Workflow for the character replacement task in video. We’ll also compare the performance between Mocha and Wan2.2 Animate, and I’ll share a few useful VRAM optimization tips to help you run smoother even on low-memory GPUs.
Step 1: Install the custom node
Install WanVideoWrapper node: https://github.com/kijai/ComfyUI-WanVideoWrapper
Search in ComfyUI Manager or instal via Git clone and pip install
Install LogicUtils node: https://github.com/aria1th/ComfyUI-LogicUtils
Search in ComfyUI Manager or instal via Git clone and pip install
Step 2: Download the model and place it in the correct folder
Place the file in:
"models/diffusion_models/MoCha/Wan2_1_mocha-14B-preview_fp8_e4m3fn_scaled_KJ.safetensors"
Place the file in: "models/loras/lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank128_bf16"
https://huggingface.co/Wan-AI/Wan2.1-T2V-14B/blob/main/Wan2.1_VAE.pth
Place the file in:
"models/vae/Wan2.1_VAE.pth"
https://huggingface.co/Kijai/WanVideo_comfy/blob/main/umt5-xxl-enc-fp8_e4m3fn.safetensors
Place the file in:
"models/text_encoders/umt5-xxl-enc-fp8_e4m3fn.safetensors"
Please check the workflow in the attachments.
Comments
Hello, 1. Is the difference only in the face area, or are other parts of the character also different compared to the reference? 2. Based on your perception, is the difference close to 100%, or roughly what percentage would you estimate? 3. Facial differences are influenced by the reference image, so please keep the face as large as possible by uploading an upper-body shot and removing as much background as you can.
SOTAI
2026-02-08 15:01:15 +0000 UTCDear SOTAI, I am wondering how keeping the face works in your video? I downloaded the workflow, changed only the reference image, video (I had to decreesed scale_to_length because of not enogh VRAM) and of course I changed the mask in order to fit to the new reference video and that's it. Everything is the same as in your workflow but in the result video the face is very apart from the reference image. As you said in the video I tried to ref2 input for Mocha embed (resized as well to be matched with the other image and wideo dimensions) but no success. The face is still totally different. Could you please advise? Thanks!
Tánczos Tibor
2026-02-08 14:47:20 +0000 UTCHi, the error occurs because the video size and the resized image size don’t match. This happens when the resolution is being changed in the wrong place. Make sure you only adjust the scale_to_length parameter in the node right after uploading the video.
SOTAI
2025-11-19 10:38:47 +0000 UTCIf I use a different resolution and framerate, I get this error: "The size of tensor a (55) must match the size of tensor b (54) at non-singleton dimension 4" what settings am I missing to allow me to use a different resolution and framerate? I'm trying to save on GPU usage.
Zach Cowan
2025-11-18 17:20:02 +0000 UTCHi, I tested it on my machine, and after installing the custom node https://github.com/aria1th/ComfyUI-LogicUtils, ComfyUI was able to detect the multiply and convert nodes. However, if this still doesn’t work on your side, you can simply remove start_times, duration nodes, multiply and convert nodes. Then set: start_index = 0 num_frame = the number of frames you want (for example, 120) inside the get_image_or_mask_from_batch node after convertToInt node. This should allow the workflow to run normally.
SOTAI
2025-11-10 02:27:53 +0000 UTCNo it’s been restarted through manager, f5, open the dev panel for windows and hard reset. Tried to install on python 13 with updated comfy independencies and tried it with the originally shipped python which I believe was 12, opened up the dev version of comfy manager found the notes that way sold them that way installed by your link they just don’t load up with comfy and are red in the workflow
Chris Rose
2025-11-10 01:05:23 +0000 UTCHi, you can find those nodes through the custom node: https://github.com/aria1th/ComfyUI-LogicUtils Sorry for the oversight, I’ve updated.
SOTAI
2025-11-07 09:03:38 +0000 UTCmultiply and convert nodes are missing, ive used dev mode in manager to install, installed from huggin and git and they will not show up
Chris Rose
2025-11-07 08:59:19 +0000 UTC
