Home Creators Posts Import Register Favorites Logout
haven't archived this post yet. have a subscription? use the importer!

Downloads

  • mocha_workflow_official_SOTAI.json

Missing 1 full-res photo, 1 file, as this post has not yet been imported.

Content

Link 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.

Please check the workflow in the attachments.

Files

Previews only

Mocha vs Wan2.2 Animate: Best Workflow for PERFECT Character Replacement in ComfyUI

Mocha vs Wan2.2 Animate: Best Workflow for PERFECT Character Replacement in ComfyUI 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. What you’ll learn: - Quick overview of Mocha and Wan2.2 Animate - Step-by-step workflow setup in ComfyUI - Performance and quality comparison - How to reduce VRAM usage and avoid crashes - Final verdict: which workflow should you choose? If you’re creating AI videos and want stable, flicker-free character replacement, this tutorial will help you understand the strengths and weaknesses of each workflow. 🔗 Related Videos: - Wan2.2 Animate workflow lowvram: [https://youtu.be/kFYxdc5PMFE] - Wan2.2 Animate workflow: [https://youtu.be/tkGVhU5hhiI] - Wan2.2 low VRAM : [https://youtu.be/WU9rr04_D4Y?si=pYVwjFfPoF6W8Kut] - ComfyUI tutorials: [ https://www.youtube.com/playlist?list=PLPb4yHX_Xwzf4oY77qD1zxW_zExbh3bXw] 📌 All Workflow: https://www.patreon.com/c/SOTAI #wananimate #comfyui #mochacomfyui

Comments

Chris Rose

multiply 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

No 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

SOTAI

Hi, 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.

Zach Cowan

If 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.

SOTAI

Hi, 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.

Tánczos Tibor

Dear 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!

SOTAI

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.