Stable Diffusion is the most capable free AI image generator available. Unlike Midjourney or DALL-E 3, it runs locally on your hardware - giving you unlimited generations, complete privacy, and total customisation.
System Requirements
- Minimum: NVIDIA GPU with 4GB VRAM (GTX 1650 or better)
- Recommended: RTX 3070 or better (8GB+ VRAM) for fast generation
- Mac: Apple Silicon Macs (M1/M2/M3) run Stable Diffusion well via Core ML
- No GPU: Use Google Colab (free) or cloud services like RunDiffusion
Installation: AUTOMATIC1111 WebUI
The most user-friendly Stable Diffusion interface. Install steps:
- Install Python 3.10.x and Git
- Clone: git clone https://github.com/AUTOMATIC1111/stable-diffusion-webui
- Download a base model (SDXL or SD 1.5) from Civitai.com
- Place it in the models/Stable-diffusion folder
- Run webui.bat (Windows) or webui.sh (Mac/Linux)
Your First Generation
Start simple: "A portrait of a young woman, soft natural lighting, photorealistic, 8k detail" - Negative prompt: "blurry, distorted, low quality, watermark"
Key Settings
- Steps: 20-30 for most cases (higher = slower but not always better)
- CFG Scale: 7-8 for balanced results (higher = more literal prompt following)
- Sampler: DPM++ 2M Karras is reliable for most prompts
- Resolution: 512x512 for SD 1.5, 1024x1024 for SDXL
Next Steps
Once comfortable with basic generation, explore: ControlNet (pose and depth control), LoRAs (style fine-tuning), inpainting (edit specific areas), and outpainting (extend image borders).
