Loading
Have a ComfyUI workflow? Get a readable Workflow Doc in your browser.Make a workflow readable.Try it now ↗
This is not just a basic Video-to-Video workflow. It features an automated First Frame Pre-Processing Pipeline that performs a head, face, body, and clothing swap on the extracted first frame of your source video...
Runtime profile
Source description
This is not just a basic Video-to-Video workflow. It features an automated First Frame Pre-Processing Pipeline that performs a head, face, body, and clothing swap on the extracted first frame of your source video before feeding it into the MiniMax H3 Reference-to-Video engine.
It allows you to seamlessly transfer motion, performance, and camera movement from a reference video while completely customizing the subject’s identity and outfit!
🔥 Key Features
Auto First-Frame Extraction & Swap: Automatically extracts the initial frame of your selected video duration and swaps the head/face/clothing with your uploaded image ( Picture 1 ) using SAM 3.1 & Flux Klein.
Dual Operating Modes: Easily toggle between full character pre-processing and direct video-to-video style transfer.
Sources
1 sourceSource excerpts
1 excerptSource context: 2504 downloads · Type Workflows · Base model MiniMax H3
⚙️ How to Use
🟢 Mode A: Full Face, Head & Clothing Swap (Default)
Load Video: Upload your reference motion video into LoadVideoUI and adjust the start/end time. (Make sure the subject's face is clearly visible in the first frame of your selection).
Upload Target Image: Place your desired character/outfit image into Picture 1 .
Switch Node: Set the Image Switch (JPS) node to Select 2 (Extracted Frame).
Run the workflow!
🔵 Mode B: Direct Video-to-Video (No Pre-Swap)
Upload your reference video into LoadVideoUI .
Upload your source image into Picture 1 .
Bypass Swap Group: Enable bypass on the FACE CLOTHE SWAP group using the Fast Groups Bypasser .
Switch Node: Set the Image Switch (JPS) node to Select 1 (Picture 1).
Run the workflow!
💬 Tips & Troubleshooting
Face Visibility: The pre-processing pipeline targets frame #1 of your selected video range. Always ensure the face is visible at the very start frame for optimal SAM 3.1 segmentation.
Do not use acceleration. You get bad results.
Custom Prompts: You can adjust the Prompt node to fine-tune the retention vs. replacement ratios if needed.
Remaining hardware or download-size estimates are not available yet.
Requirements
Checkpoint · 8.91 GB · Hugging Face · black-forest-labs/FLUX.2-klein-base-9b-fp8
VAE · Hugging Face · Comfy-Org/flux2-dev
Minimax ref2v turbo 4step_v0.1_comfyui_bf16.safetensors
Not resolvedLoRA · Unknown
minimax_h3_audio_vae_fp32.safetensors
File unverifiedVAE · Hugging Face · Comfy-Org/MiniMax-H3 · MiniMax H3
minimax_h3_ref2va_pruned_int8_convrot.safetensors
File unverifiedCheckpoint · Hugging Face · Comfy-Org/MiniMax-H3 · MiniMax H3
minimax_h3_video_vae_fp16.safetensors
File unverifiedVAE · Hugging Face · Comfy-Org/MiniMax-H3 · MiniMax H3
qwen_3_8b_fp8mixed.safetensors
Not resolvedText encoder · Hugging Face · Comfy-Org/flux2-klein-9B
qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors
File unverifiedText encoder · Hugging Face · Comfy-Org/MiniMax-H3 · MiniMax H3
sam3.1_multiplex_fp16.safetensors
Not resolvedVision model · Unknown
Node pack · Registry
comfyui-easy-use
FoundNode pack · Registry
comfyui-impact-pack
FoundNode pack · Registry
Node pack · Registry
comfyui-videohelpersuite
FoundNode pack · Registry
comfyui-workflow-encrypt
FoundNode pack · Registry
Node pack · Registry
