gguf-comfy 0.1.9

run flux1/sd3 model with beginner GPU (low cost) or even CPU

gguf-comfy *this project will gradually migrate to gguf-node

custom node diagram generator (image/audio/video)

download the portal via

py -m gguf_comfy

• decompress the 7z bundle file: Extract All... (it includes everything you need to run a model)
• you could either get the dry run pack here or pick them one-by-one from the original source below
• download flux1-dev-Q4_0.gguf (6.32GB); pull it to ./ComfyUI/models/unet
• download clip_l.safetensors (234MB) and t5xxl_fp8_e4m3fn.safetensors (4.55GB); pull them to ./ComfyUI/models/clip
• download ae.safetensors (319MB); pull it to ./ComfyUI/models/vae
• run the .bat file under the main directory (it will activate the py backend as well as the js frontend)
• drag gguf-workflow.json to the activated browser

You are good to GO! (now you can run flux1 with the cheapest Nvidia GPU or merely CPU) ENJOY!

Reference: flux comfyanonymous city96

Stable Diffusion 3.5 is now supported (please upgrade to the latest version)

Run sd3.x model following the step-by-step guide below:

• you could either get the backup pack here or pick them one-by-one from the original source below
• download sd3.5_large-q4_0.gguf (4.77GB); pull it to ./ComfyUI/models/unet
• download clip_g.safetensors (1.39GB), clip_l.safetensors (234MB), and t5xxl_fp8_e4m3fn.safetensors (4.55GB); pull them all to ./ComfyUI/models/clip
• download diffusion_pytorch_model.safetensors (168MB); pull it to ./ComfyUI/models/vae
• run the .bat file under the main directory (it will activate the py backend as well as the js frontend)
• drag workflow-sd3.5-gguf.json to the activated browser

Reference: stability-ai gguf