Viggle Remake - EASY (ComfyUI workflow)
One ComfyUI workflow for Swap (replace a person or character) and Edit (change objects, background, weather, light). Any video length from 1 s to 18 s, with an optional upscale to 1080p. Built on MiniMax H3 and the Viggle Remake LoRAs.
Files in this repo
| file | what it is |
|---|---|
| Viggle Remake - EASY HQ (Swap + Edit, any length).json | 16 GB+ VRAM workflow - drag it into ComfyUI |
| Viggle Remake - EASY 12GB VRAM (Swap + Edit, any length).json | 12 GB VRAM workflow - drag it into ComfyUI |
| viggle-easy-custom-nodes.zip | 4 custom node packs both workflows need - unzip into ComfyUI/custom_nodes/ |
Which workflow should I use?
| HQ (16 GB+) | 12 GB VRAM | |
|---|---|---|
| graphics card | 16 GB VRAM or more | 12 GB VRAM |
| system RAM | 32 GB+ | 32 GB minimum, 64 GB recommended |
| video model | MiniMax H3 int8 + Viggle LoRAs | Viggle GGUF Q8_0 (LoRA merged in); Q4_K_M for low RAM |
| extra node pack | - | ComfyUI-GGUF |
| ComfyUI start flag | none | --reserve-vram 8 (Desktop: Settings > Server-Config > Reserved VRAM = 8) |
| resolution | 480 (raise to 576-736 for a sharper face) | 480 (832x480) - keep it there |
| upscale tiling | off (faster) | on (saves VRAM) |
| speed | faster | slower (the model streams from system RAM) |
Swap/Edit, any length, the upscale to 1080p, and how you use it are the same in both. Each workflow has its own DOWNLOADS note on the left with exactly the files it needs. The 12 GB settings follow the Viggle Remake card table; its 1080p upscale has not been tested on a real 12 GB card.
What you need (HQ workflow)
The workflow has a DOWNLOADS note on the left with every link. The same list.
12 GB: replace the first three files below with viggle_remake_swap-Q8_0.gguf + viggle_remake_edit-Q8_0.gguf (22.3 GB each, models/diffusion_models/),
or the low-RAM swap-Q4_K_M + edit-Q4_K_M (14.3 GB each).
Models - every run
| file | put it in | size |
|---|---|---|
| minimax_h3_fl2va_pruned_int8_convrot.safetensors | models/diffusion_models/ |
21.0 GB |
| viggle_remake_swap_comfyui.safetensors | models/loras/ |
3.8 GB |
| viggle_remake_edit_comfyui.safetensors | models/loras/ |
3.8 GB |
| qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors | models/text_encoders/ |
15.7 GB |
| minimax_h3_video_vae_fp16.safetensors | models/vae/ |
5.2 GB |
| minimax_h3_audio_vae_fp32.safetensors | models/vae/ |
0.61 GB |
Models - only for UPSCALE to 1080p
| file | put it in | size |
|---|---|---|
| minimax_h3_ref2va_pruned_int8_convrot.safetensors | models/diffusion_models/ |
21.0 GB |
| minimax_h3_taomate_3step_lora_avg_rank_19_bf16.safetensors | models/loras/ |
0.18 GB |
| h3-realism-people-t2v-i2v-r2v.safetensors | models/loras/ |
0.13 GB |
| minimax_h3_latent_upscaler_3d_fp16.safetensors | models/latent_upscale_models/ |
0.69 GB |
Rename the last one to minimax_h3_latent_upscaler_3d_fp16.safetensors after downloading.
Custom nodes (into ComfyUI/custom_nodes/, then restart ComfyUI)
- ComfyUI-KJNodes
- comfyui-minimax-h3-audio-T8
- ComfyUI-H3-FaceRefine (no pip install needed)
- ComfyUI-Viggle-Remake - folder
custom_nodes/ComfyUI-Viggle-Remake/with init.py and nodes.py - viggle-easy-custom-nodes.zip - ComfyUI-Viggle-Easy, pika-h3-lowram, ComfyUI-H3-UltimateUpscale-LD, ComfyUI-H3-UpscaleSize
ComfyUI 0.44.2 or newer. Made for a 16 GB+ NVIDIA card and 32 GB+ system RAM.
For AI agents
Download the workflow JSON and read extra.viggle_downloads: every model and node pack with its URL, target path, size and sha256,
plus step-by-step install instructions. Each loader node also carries ComfyUI's properties.models download hints.
Credits and licenses
This repo holds only the workflow and small helper nodes; the models stay with their authors and their licenses: MiniMax H3 (Comfy-Org repack), Viggle Remake, TaoMate LoRA (Kijai), realism LoRA (fal), 3D latent upscaler (LBH-123-AI). ComfyUI-H3-UltimateUpscale-LD and ComfyUI-H3-UpscaleSize are based on PlagueKind's nodes (MIT, license included in the zip). Only use real people's faces with their permission.
- Downloads last month
- -