Wan2.1 GGUF β€” 4GB VRAM ComfyUI Workflow

By: The_frizzy1 Hardware target: RTX 3050 Laptop (4 GB VRAM) CivitAI: https://civitai.com/models/1309674/wan21gguf-only-4gb-vram-comfyui-workflow YouTube: https://www.youtube.com/@the_frizzy1


What This Is

A ComfyUI workflow pack for running Wan2.1 video generation on low-VRAM GPUs (as low as 4 GB). Uses GGUF-quantised models and custom nodes from Kijai and others.


Included Workflows

  • Wan2.1 T2V β€” Text to Video
  • Wan2.1 I2V β€” Image to Video
  • Wan2.1 VACE β€” Video and Conditioning Extension
  • Wan2.1 First Frame / Last Frame
  • Funcontrol (experimental)
  • Funcameraimage (experimental)

Required Custom Nodes


Model Downloads

VAE / CLIP / Text Encoder: https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files

Place all .gguf files in ComfyUI/models/diffusion_models/


Quantisation Guide

Level VRAM Quality
Q5 Medium Best balance
Q3_K_M Low Good
Q2_K Lowest Usable
14B High Best quality

Performance Tips

Enable Xformers, Sage Attention, or Triton for extra speed. Triton install guide: https://www.patreon.com/posts/easy-guide-sage-124253103


Helpful Links


Changelog

Version Notes
v2.0 Teacache, Torch Compile, new GGUF loader
v1.8 See install video
v1.1 Better quality and stability
v1.0 Initial release
Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. πŸ™‹ Ask for provider support

Model tree for The-frizzy1/Wan21-GGUF-4GB-Workflow

Finetuned
(54)
this model