Coercer/ChuckFnS_GGUF

Chuck Fn'S Model, GGUF version to run on low VRAM/RAM builds. Done with: https://colab.research.google.com/drive/1xRwSht2tc82O8jrQQG4cl5LH1xdhiCyn?usp=sharing

Go to your ComfyUI 'models' folder, and drag/drop the files in their respective folders.

You need the following custom node to load GGUF models: https://github.com/city96/ComfyUI-GGUF (Install manually or via ComfyUI Manager).

Then, load this workflow (Right click > Download Link, then drag and drop into ComfyUI): https://huggingface.co/Coercer/ChuckFnS_GGUF/resolve/main/GGUF_Workflow.json

Requirements, from higher RAM to lower RAM:

Normal .safetensors > .safetensors at fp8 = Q8_0 > Q6_K > Q5_K_M > Q5_K_S > Q4_K_M > Q4_K_S > Q3_K_L > Q3_K_M > Q3_K_S > Q2_K

(Probably, the lowest ones have reduced quality, although that's for you to check in your own prompts. The agreed minimum acceptable quality is said to be Q4_K_M, but it ultimately depends on the style you're aimimng at).

If you REALLY need to save VRAM, use the modular workflow. You'll need this custom node to run it: https://github.com/endman100/ComfyUI-SaveAndLoadPromptCondition.git (Use install via GIT Url option). Further instructions are present in the workflows.