GitHub - Martynienas/FooocusPocus: Focus on prompting and generating

FooocusPocus

A Quality-of-Life Enhanced Fork of Fooocus

>>> Click Here to Install FooocusPocus <<<

About This Fork

FooocusPocus is a fork of Fooocus by lllyasviel.

Goal: keep core Fooocus generation behavior while improving usability, workflow, and configuration quality of life.

For full upstream capability docs, see the original Fooocus repository.


Highlights

Image Library (Major Feature)

The Image Library is an in-app browser for generated images with metadata-aware management.

  • Browse generated images in a dedicated modal gallery
  • Search images by prompt text
  • Filter by tags
  • Single-image preview with full metadata panel
  • Multi-select with checkboxes for bulk operations
  • Unified delete action:
    • Deletes one image when single-selected
    • Deletes multiple images when checkbox selection is active
  • Edit tags for one or many images
  • Optional auto-reload when opening the library

Configuration Tab

A dedicated tab for managing runtime settings without manual config editing.

  • Add/remove checkpoint folders and LoRA folders
  • Auto-reload models after folder changes
  • Edit key paths (output/temp/embeddings/VAE/ControlNet/upscale)
  • Set generation defaults (steps/CFG/sampler/scheduler/model/style/aspect ratio/output format)
  • Per-setting reset buttons and full restore-to-defaults
  • Auto-save configuration changes

UI and Workflow Improvements

  • Prompts are organized for faster day-to-day use
  • Better operation feedback and console logging
  • Random LoRA option for exploration

Prompt Syntax Reference

FooocusPocus supports both wildcard placeholders and dynamic prompt groups.

1) Wildcard files (__name__)

Use wildcard placeholders in prompts:

  • a portrait of __artist__
  • __color__ sports car in __city__

Wildcard files are loaded from your configured wildcards folder (path_wildcards), with one option per line.

2) Dynamic choices ({...})

Use inline dynamic groups:

  • Single choice: {red|green|blue}
  • Single choice (spaces are fine): {red | green | blue}
  • Multi-select count: {2$$red|green|blue|yellow}
  • Multi-select range: {1-3$$red|green|blue|yellow}

Both positive and negative prompts support this syntax.

3) Read wildcards in order

When enabled, wildcard file entries are consumed deterministically by index (useful for reproducible batches).
When disabled, wildcard entries are chosen randomly.


Download

Windows

You can download FooocusPocus from the Releases page.

After downloading, extract and run run.bat.

System Requirements

  • Minimum: 4GB Nvidia GPU VRAM and 8GB system RAM
  • Recommended: 6GB+ VRAM and 16GB+ RAM

Changes from Upstream (Summary)

Feature Description
Image Library In-app generated image browser with metadata, tag filter/search, multiselect, bulk delete, and tag editing
Configuration Tab UI-based settings management with auto-save and reset controls
Dynamic Model Folders Add/remove checkpoint and LoRA folders without restart
Prompt Utilities Wildcard placeholders and dynamic prompt groups
Random LoRA Optional random LoRA selection for experiments
UX/Logging Improved feedback and operational visibility

Original Fooocus Features

FooocusPocus includes upstream Fooocus capabilities, including:

  • High-quality text-to-image generation
  • Inpainting and outpainting
  • Image prompt workflows
  • Style systems and model switching
  • Upscale and variation workflows

For full documentation, see Fooocus upstream.


Contributing

Contributions are welcome via pull requests and issues.

License

This project inherits Fooocus licensing. See LICENSE.

Credits

  • Original Fooocus by lllyasviel
  • FooocusPocus enhancements by contributors

Note: This is an unofficial fork. For the official project, visit github.com/lllyasviel/Fooocus.