Infinirc
安裝開源FLUX.1-dev搭配ComfyUI
安裝開源FLUX.1-dev搭配ComfyUI

FLUX.1是AI圖像生成領域的重大突破,由Black Forest Labs團隊開發的這款開源模型,在圖像細節、構圖和美感方面已超越了許多主流模型。本教學將引導您在ComfyUI環境中安裝並運行FLUX.1-dev,為您的創意項目提供頂級AI圖像生成能力。

發佈於  星期一,三月 17 2025

概述

透過本指南,您將學習:

  • 如何識別並選擇適合您設備的FLUX.1版本
  • 獲取並正確配置所需檔案的完整流程
  • 建立基礎工作流程以快速開始使用模型
  • 優化設置以獲得最佳性能和圖像質量
版本特性輕量級版本平衡版本完整高清版本
模型名稱FLUX.1單檔案FP8FLUX.1 SchnellFLUX.1-dev FP16
總檔案大小約17GB約23GB約34GB
核心模型檔案單一整合檔案分離式結構完整參數結構
所需記憶體最低16GB VRAM最低16GB VRAM建議24GB+ VRAM
運算速度中等最快需較長處理時間
圖像質量良好優良最佳效果
適用情境入門使用或硬體受限日常創作需求專業作品與高要求專案
安裝複雜度簡單中等較複雜

安裝ComfyUI

git clone https://github.com/comfyanonymous/ComfyUI.git
cd ComfyUI
pip install -r requirements.txt
python main.py

安裝FLUX.1單檔案FP8

flux1-schnell-fp8.safetensors

https://huggingface.co/Comfy-Org/flux1-dev/blob/main/flux1-dev-fp8.safetensors
將下載的檔案放到ComfyUI/models/unet/

安裝FLUX.1 Schnell

flux1-schnell-fp8.safetensors

https://huggingface.co/Comfy-Org/flux1-schnell/blob/main/flux1-schnell-fp8.safetensors
將下載的檔案放到ComfyUI/models/unet/

t5xxl_fp8_e4m3fn.safetensors

https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors
將下載的檔案放到ComfyUI/models/CLIP/

ae.safetensors

https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors
將下載的檔案放到ComfyUI/models/VAE/

clip_l.safetensors

https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors
將下載的檔案放到ComfyUI/models/CLIP/

安裝FLUX.1-dev FP16

flux1-dev.safetensors

https://huggingface.co/black-forest-labs/FLUX.1-dev/resolve/main/flux1-dev.safetensors?download=true
將下載的檔案放到ComfyUI/models/unet/

t5xxl_fp8_e4m3fn.safetensors

https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors
將下載的檔案放到ComfyUI/models/CLIP/

ae.safetensors

https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors
將下載的檔案放到ComfyUI/models/VAE/

clip_l.safetensors

https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/clip_l.safetensors
將下載的檔案放到ComfyUI/models/CLIP/

  • Artificial-Intelligence