Fooocus, a modern interface for Stable Diffusion XL, offers advanced features like image prompting. This powerful tool allows you to use images alongside text prompts to guide AI image generation, enabling style replication, composition control, and even face swapping.
Understanding Image Prompts
Unlike image-to-image generation, image prompts complement text prompts, influencing the entire image creation process. Fooocus uses advanced techniques, including:
- IP-Adapter integration
- Negative embedding
- Attention hacking
- Adaptive weighting algorithms
These enhancements make Fooocus' image prompting more effective than standard implementations.
Getting Started with Image Prompts
- Activate "Image Input" below the text prompt field
- Click on the "Image Prompt" tab
- Upload up to 4 reference images
Basic Image Prompting
Let's compare a standard text-to-image generation with an image-prompted one:
- Text-only prompt: "full body shot of a Beautiful girl, smiling, Urban Chic, Edgy, New York City, spring, blooming flowers, Central Park, iconic skyline, Empire State and Chrysler Buildings"
- Same text prompt + image prompt (e.g., 1woman, rooftop portrait)
Advanced Image Prompt Settings
Adjusting Image Influence
Two key parameters control image prompt strength:
- Weight: Determines the image's influence (default: 0.6)
- Stop At: Sets when the image stops influencing generation (default: 0.5)
Multi-Image Prompting
Combine multiple reference images for creative results:
ControlNet in Fooocus
Fooocus implements ControlNet-like features through two modes:
- PyraCanny: Edge detection optimized for high-resolution images
- CPDS: Structure recognition based on Contrast Preserving Decolorization
Face Swapping with FaceSwap
The FaceSwap mode allows you to incorporate specific facial features:
- Upload an image with a clear face
- Adjust Weight and Stop At for desired effect
- Combine with text prompts for varied results
Combining Techniques
For ultimate control, mix different image prompt modes:
Top comments (0)