Professional-grade image generation that runs entirely on your own GPU. ComfyUI + FLUX.1 Dev gives you Midjourney-quality output with full creative control and zero data leaving your machine.
What You Need
- A GPU with 12GB+ VRAM (24GB recommended)
- Python 3.10+ or the ComfyUI desktop app
- About 20 minutes
Setup
Option A: Desktop App (Easiest)
Download from comfy.org, install, and use the built-in model manager to download FLUX.1 Dev.
Option B: Manual Install
git clone https://github.com/Comfy-Org/ComfyUI.git
cd ComfyUI
pip install -r requirements.txt
python main.py
Open http://localhost:8188.
Basic FLUX Workflow
- Add a Checkpoint Loader node - load flux1-dev.safetensors
- Add CLIP Text Encoder - enter your prompt
- Add KSampler - connect model, CLIP, and empty latent
- Add VAE Decode - decode to image
- Add Save Image - save result
- Click Queue Prompt
Prompt example: "a photorealistic cat sitting on a vintage leather chair, warm lighting, depth of field"
Advanced Features
- LoRA - add a LoRA Loader node for style control
- ControlNet - pose/edge guidance with extra nodes
- Image-to-Image - feed an existing image through VAE Encode
- API mode - integrate with n8n or custom apps
- Batch generation - queue multiple prompts at once
Cost vs Cloud
| Local | Midjourney | |
|---|---|---|
| Monthly | $0 | $10-60 |
| Per image | $0 | $0.04-0.12 |
| Privacy | Stays on your GPU | Sent to cloud |
| Control | Full node-level | Limited |
Full guide with troubleshooting and hardware tips: https://everylocalai.com/stack/comfyui-flux-local-image
Top comments (0)