Fooocus forks and variants

Fri 05 January 2024

Similar to other popular open source AI projects like AUTOMATIC1111, Fooocus has many community-maintained forks. Below is a non-comprehensive list of all the popular ones:

Each and every fork has their own set of improvements and new features.

Fooocus vs Fooocus-Control

Fooocus-Control is now renamed to Fooocus-ControlNet-SDXL. The software is a fork of Fooocus combined with ControlNet, SDXL and applied a IP-Adapter to add advanced features: - Sketch: Use a sketch as image prompt input and generate images with elements in the sketch included. - Depth: "Copy" the depth of one image over another. - Pose: "Copy" the pose from image input and generate variations. - ReColor: Repaint the color into the imput image. - MaskInpaint: Create a mask and lets the AI do its thing in that area only. Fooocus-ControlNet-SDXL Sketch feature

There are more ControlNet features coming, see more at https://github.com/fenneishi/Fooocus-ControlNet-SDXL

Fooocus vs RuinedFooocus

RuinedFooocus is another Fooocus fork. Multiple enhancements have been added, mostly improvements on user interface.

Customization and Personalization

  • Custom Styles and Themes: Select from custom styles and themes in settings/styles.csv.
  • Custom Resolutions: Add and select custom resolutions.
  • Performance Settings: Customize samplers, steps, and advanced performance settings.
  • ControlNet and PowerUp: Advanced image manipulation and custom default settings.

User Interface Enhancements

  • Dropdown Menus: Use dropdowns for selecting styles and resolutions.
  • Metadata Management: Save and view full metadata for images.
  • UI Defaults: Set and change default UI settings.
  • Readable Display: Enhanced readability for resolution modes.

Image Generation Options

  • Multiple Styles Application: Apply multiple styles to a single prompt.
  • Random Prompt Generation: Generate random prompts with a dedicated tab.
  • Batch Processing: Process a list of prompts at once.
  • Image2Image Mode: Transform existing images into new creations.
  • Inpainting and Evolve Features: Fill areas and evolve images into variations.

Workflow and Productivity

  • Drag-and-Drop Functionality: Drag images to autofill prompts.
  • Prompt Enhancements: Enhance prompts with loras and style directives.
  • Shortcut Keys: Quick generation with keyboard shortcuts.
  • Cancel Button: Terminate image generation when needed.

Model and Format Support

  • Model Compatibility: Supports SSD-1B Models and SDXL LCM LORA.
  • Upscaling: Improve image resolution with upscaling options.

Automation and Efficiency

  • Automatic Negative Prompts: Create negative prompts automatically.
  • Automatic Triggerword Downloads: Fetch Lora triggerwords without manual effort.
  • Generate Forever Mode: Unlimited image generation.

Security and Updates

  • Basic Security: Simple username/password protection.
  • Software Upgrades: Easy updates for xformers and torch.

Accessibility and Notifications

  • No Browser Mode: Use the app without launching a web browser with --nobrowser flag.
  • Notification System: Audio alerts upon generation completion.
  • Metadata Viewer: Easily inspect image metadata.

More info: https://github.com/runew0lf/RuinedFooocus

Fooocus vs Fooocus-MRE

Fooocus-MRE stands for "MoonRide Edition", named after its main developer/maintainer MoonRide303. It's got a few more features compared to the original Fooocus: - Image-2-Image: Support for Image-2-Image mode, Upscaling via Image-2-Image (requires more VRAM), generate multiple images using same seed in Image-2-Image mode. - Control-LoRA model card: Canny Edge, Revision, Revision are supported. - Limited support for non-SDXL models (no refiner, Control-LoRAs, Revision, inpainting, outpainting). - Full metadata display for generated images - Support for FreeU (a method to substantially improves diffusion model sample quality at no additional cost). - Metadata saving (JSON or embedded) and loads prompt information from JSON/image files - Generate images forever mode - More compact resolution & style selection - Audio notification on generation finish - Hotkey generation added (Ctrl+ENTER) - Support for loading models from subfolders (ported from RuinedFooocus) - Authentication supported, useful in --share mode. - Style Iterator (iterates over selected style(s) combined with remaining styles - S1, S1 + S2, S1 + S3, S1 + S4, and so on; for comparing styles pick no initial style, and use same seed for all images). - Limited support for non-SDXL models

More info: https://github.com/MoonRide303/Fooocus-MRE

Fooocus vs Fooocus-SAM

SAM stands for Segment-Anything Model, which primarily aims to produce high quality object masks from input prompts such as points or boxes. See the demo image and you'll know how it works: Having marked an area as a mask, we can keep or reproduce Fooocus-SAM adds EfficientSAM, which is a compact variant of Segment-Anything Model to Fooocus. It is solely the biggest improvement of this project. Let's see what we can do with it: Fooocus-SAM draw the mask

Fooocus-SAM mask identified More information: https://github.com/pppoe/Fooocus-SAM

Fooocus vs SimpleSDXL

SimpleSDXL is a fork of Fooocus aimed towards Chinese users. Its main features includes: - Chinese and English prompts supported. - Chinese UI (texts and buttons) - Improved image gallery browsing. - Local model source is supported (to avoid the Chinese GFW) - Segment-Anything Model supported for masking.

More info: https://github.com/metercai/SimpleSDXL

Category: misc