SmoothMIX WAN 2.2 Finetune - ComfyUI & One Click Installer

SmoothMIX WAN 2.2 Finetune introduces a refined motion and animation profile over the base WAN 2.2, delivering smoother transitions and more responsive timelines with lower latency. It emphasizes efficiency with a streamlined finetuning approach that preserves your core WAN 2.2 capabilities while sharpening motion fidelity and stability across sequences. In short, you get faster, more natural-looking motion without needing LoRA layers or complex retraining.
 
Preloaded Models Within the Installer (Low VRAM)
umt5_xxl_fp8_e4m3fn_scaled.safetensors (clip folder) — Hugging Face
wan_2.1_vae.safetensors (vae folder) — Hugging Face
smoothMixWan22I2VT2V_i2vHigh-Q4_K_M.gguf (unet folder) — Hugging Face
smoothMixWan22I2VT2V_i2vLow-Q4_K_M.gguf (unet folder) — Hugging Face
clip_vision_h.safetensors (clip_vision folder) — Hugging Face
2xLexicaRRDBNet_Sharp.pth upscale model — Hugging Face
Note: The standard high-fidelity SmoothMIX Wan2.2 diffusion models (FP16 and FP8) are not included in the installer for size reasons. For best results, download them separately from the CivitAI model page.
 
Speed
Generate 5-second 480p videos (6 steps) in less than a minute on an RTX 4090 24 GB VRAM GPU. The workflow dynamically adapts — faster render times and higher resolutions (720p/24fps) are achievable on higher-end GPUs.
 
System Requirements
Nvidia RTX 30XX, 40XX, or 50XX series GPU (FP16 support required; GTX 10XX/20XX not tested)
CUDA-compatible GPU (minimum 6 GB VRAM; more recommended)
Windows OS
At least 40 GB free storage
 
What's Included
Portable, pre-configured ComfyUI and SmoothMix Wan2.2 install
Automated download of all custom nodes and required models
Workflow designed for simplicity and speed with modular Fast Groups Bypasser
 
Usage Notes
You can utilize either the SmoothMix Wan2.2 GGUF models or the standard SmoothMix Wan2.2 diffusion models.
Enable or disable workflow sections using the levers in the Fast Groups Bypasser node on the left side of the workflow.
Enter your detailed descriptive text prompt for the scene you want to generate.
For best results, enhance your prompt with an LLM.