README.md

---
license: mit
language:
- en
tags:
- comfyui
- gguf-comfy
- gguf-node
widget:
- text: >-
    anime style anime girl with massive fennec ears and one big fluffy tail, she
    has blonde long hair blue eyes wearing a maid outfit with a long black gold
    leaf pattern dress, walking slowly to the front with sweetie smile, holding
    a fancy black forest cake with candles on top in the kitchen of an old dark
    Victorian mansion lit by candlelight with a bright window to the foggy
    forest
  output:
    url: samples\ComfyUI_00003_.webp
- text: drag it <metadata inside>
  output:
    url: samples\ComfyUI_00007_.webp
- text: drag it <metadata inside>
  output:
    url: samples\ComfyUI_00001_.webp
---

# **gguf-node test pack**

![screenshot](https://raw.githubusercontent.com/calcuis/comfy/master/gguf-node.png)
locate **gguf** from Add Node > extension dropdown menu (between 3d and api; second last option)

[<img src="https://raw.githubusercontent.com/calcuis/comfy/master/gguf.gif" width="128" height="128">](https://github.com/calcuis/gguf)

### **setup (in general)**
- drag gguf file(s) to diffusion_models folder (`./ComfyUI/models/diffusion_models`)
- drag clip or encoder(s) to text_encoders folder (`./ComfyUI/models/text_encoders`)
- drag controlnet adapter(s), if any, to controlnet folder (`./ComfyUI/models/controlnet`)
- drag lora adapter(s), if any, to loras folder (`./ComfyUI/models/loras`)
- drag vae decoder(s) to vae folder (`./ComfyUI/models/vae`)

## **run it straight (no installation needed way; recommended)**
- get the comfy pack with the new gguf-node ([pack](https://github.com/calcuis/gguf/releases))
- run the .bat file in the main directory

## **or, for existing user (alternative method)**
- you could git clone the node to your `./ComfyUI/custom_nodes` (more details [here](https://github.com/calcuis/gguf/))
- either navigate to `./ComfyUI/custom_nodes` first or drag and drop the node clone (gguf repo) there

### **workflow**
- drag any workflow json file to the activated browser; or
- drag any generated output file (i.e., picture, video, etc.; which contains the workflow metadata) to the activated browser

### **simulator**
- design your own prompt; or
- generate random prompt/descriptor(s) by the [simulator](https://prompt.calcuis.us) (might not applicable for all models)

<Gallery />

#### **booster**
- drag safetensors file(s) to diffusion_models folder (./ComfyUI/models/diffusion_models)
- select the safetensors model; click `Queue` (run); simply track the progress from console
- when it was done; the boosted safetensors fp32 will be saved to the output folder (./ComfyUI/output)

#### **cutter (beta)**
- drag safetensors file(s) to diffusion_models folder (./ComfyUI/models/diffusion_models)
- select the safetensors model; click `Queue` (run); simply track the progress from console
- when it was done; the half-cut safetensors fp8_e4m3fn will be saved to the output folder (./ComfyUI/output)

### **convertor (alpha)**
- drag safetensors file(s) to diffusion_models folder (./ComfyUI/models/diffusion_models)
- select the safetensors model; click `Queue` (run); track the progress from console
- the converted gguf file will be saved to the output folder (./ComfyUI/output)

### **convertor (reverse)**
- drag gguf file(s) to diffusion_models folder (./ComfyUI/models/diffusion_models)
- select the gguf model; click `Queue` (run); track the progress from console
- the reverse converted safetensors file will be saved to the output folder (./ComfyUI/output)

### **convertor (zero)**
- drag safetensors file(s) to diffusion_models folder (./ComfyUI/models/diffusion_models)
- select the safetensors model; click `Queue` (run); track the progress from console
- the converted gguf file will be saved to the output folder (./ComfyUI/output)
- zero means **no restrictions**; different from alpha; any form of safetensors can be converted; pig architecture will be applied for the output gguf

### **latest feature: gguf vae🐷 loader is now working on gguf-node**
- gguf vae is able to save memory consumption of your machine
- convert your safetensors vae to gguf vae using **convertor (zero)**
- then use it with the new **gguf vae loader**
- same as gguf clip loaders, gguf vae loader is compatible with both safetensors and gguf file(s)

### **disclaimer**
- some models (original files) as well as part of the codes are obtained from somewhere or provided by someone else and we might not easily spot out the creator/contributor(s) behind, unless it was specified in the source; rather let it blank instead of anonymous/unnamed/unknown
- we hope we are able to make more effort to trace the source; if it is your work, do let us know; we will address it back properly and probably; thanks for everything

### **reference**
- sd3.5, sdxl from [stabilityai](https://huggingface.co/stabilityai)
- flux from [black-forest-labs](https://huggingface.co/black-forest-labs)
- pixart from [pixart-alpha](https://huggingface.co/PixArt-alpha)
- lumina from [alpha-vllm](https://huggingface.co/Alpha-VLLM)
- aura from [fal](https://huggingface.co/fal)
- mochi from [genmo](https://huggingface.co/genmo)
- hyvid from [tencent](https://huggingface.co/tencent)
- wan from [wan-ai](https://huggingface.co/Wan-AI)
- ltxv from [lightricks](https://huggingface.co/Lightricks)
- cosmos from [nvidia](https://huggingface.co/nvidia)
- pig architecture from [connector](https://huggingface.co/connector)
- comfyui from [comfyanonymous](https://github.com/comfyanonymous/ComfyUI)
- comfyui-gguf from [city96](https://github.com/city96/ComfyUI-GGUF)
- llama.cpp from [ggerganov](https://github.com/ggerganov/llama.cpp)
- llama-cpp-python from [abetlen](https://github.com/abetlen/llama-cpp-python)
- gguf-connector ([pypi](https://pypi.org/project/gguf-connector)|[repo](https://github.com/calcuis/gguf-connector))
- gguf-node ([pypi](https://pypi.org/project/gguf-node)|[repo](https://github.com/calcuis/gguf)|[pack](https://github.com/calcuis/gguf/releases))