wan-gguf

wan-gguf

GGUF quantized versions of the Wan video models for text-to-video work in ComfyUI or diffusers, including small 1.3b files for older machines. For people who make short AI video locally.

Made by calcuis on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face.

About

What it is: wan-gguf is a collection of GGUF quantized versions of the Wan video model, repackaged from the ComfyUI release of Wan 2.1. You use the files to make short videos from a text prompt on your own computer. The 1.3b text-to-video and VACE files are noted as working well on older or low-end machines.

Who it suits: It suits people who generate AI video locally and want smaller model files that run on modest hardware. It also suits tinkerers who already work in ComfyUI or diffusers.

Title: Who uses it
Industry | Who uses it | How they use it | Example
Video and content | Creators | Make short clips from a text prompt | Generate a clip from a scene description
Software | Tool developers | Run the model through diffusers in a script | Export a generated video to a file
Hobby use | Owners of older machines | Use the 1.3b GGUF files on low-end hardware | Run text-to-video on an aging computer

People who want a hosted video service can skip it.

Where to use it:

• ComfyUI, with a GGUF node such as gguf-node
• Python with the diffusers library
• The gguf-connector command line
• Windows, macOS or Linux

Title: Top jobs
Use case | Who | What they get back
Text-to-video | Creators | A short generated clip from a prompt
Image-to-video | Creators | A clip that starts from an image, using the extra vision file
VACE editing | Developers | A video from the 1.3b VACE file through diffusers

How to set it up:

• Python on your computer
• The huggingface_hub package
• ComfyUI with a GGUF node, or diffusers
• Enough memory for the file you choose

1. Install the Hugging Face hub package.
2. Download the files to your computer.
3. Put each file in its ComfyUI folder, as in the table below.
4. Load the workflow and run it.

pip install -U huggingface_hub
hf download calcuis/wan-gguf

Title: Where the files go
File | Folder
GGUF model | ./ComfyUI/models/diffusion_models
t5xxl-um text encoder | ./ComfyUI/models/text_encoders
VAE | ./ComfyUI/models/vae
clip-vision-h, for image-to-video only | ./ComfyUI/models/clip_vision

How to use it: In ComfyUI, drag a demo video from the model page into your browser to load a workflow. If you use the author's gguf pack, run the .bat file in its main directory. With the fp8 scaled text encoder, set the device in the GGUF clip loader to cpu, which the card says does not affect speed. The other route is the gguf-connector command.

ggc v2

Example: In the diffusers route, the card prompts for a short clip of a pig moving fast through a snowy forest at sunset, with a tracking camera. It sets the size to 720 by 480, 57 frames and 24 steps, then exports the result at 16 frames a second. The card shows no sample video here.

For newcomers and experts: If you are new, start with ComfyUI and one demo workflow from the model page. Try the small 1.3b text-to-video file first, since it is the one noted for modest machines.

For experts, the pig architecture is a loose label that the gguf node accepts for any model, encoder or VAE file. In the comfyui-gguf node you may need to add pig to its image architecture list in loader.py. The newest 1.3b VACE file keeps everything except block weights in f32 to avoid embedding key errors.

Download the GGUF, encoder and VAE files
Place them in the ComfyUI folders
Load a demo workflow or the diffusers script
Generate and export the video

Good to know:

• Pig label: the architecture name needs support in your GGUF node.
• fp8 encoders: use cpu offload in the clip loader, per the card.
• License: Apache 2.0.
• Cost to run: none named.
• Data it reaches: it downloads files from Hugging Face, and the source names nothing else.
• Not for: anyone who wants a one-click app, since setup takes several files.

On Hugging Face, 13 November 2026

• Task: text to video
• Parameters: 127M
• Weights: about 63 MB at 4-bit (estimate)
• Formats: GGUF
• Downloads in the last 30 days: 21,309
• Likes: 210
• License: apache-2.0
• Languages: en
• Based on: Comfy-Org/Wan_2.1_ComfyUI_repackaged
• Tags: gguf, gguf-connector, gguf-node, text-to-video

Links

• Model: https://huggingface.co/calcuis/wan-gguf
• Files: https://huggingface.co/calcuis/wan-gguf/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download calcuis/wan-gguf

Then follow the usage in its model card. Nothing in this kit runs by itself.