Customizing Fooocus

Configuration Files

The main configuration file is config.txt, located in the Fooocus root directory. This file allows you to set default values and customize behavior.

Basic Configuration Structure

# Model Paths
base_model_path = models/checkpoints/
lora_path = models/loras/
vae_path = models/vae/

# Default Parameters
default_steps = 30
default_cfg_scale = 7.0
default_sampler = dpmpp_2m_sde_gpu

# Output Settings
output_path = outputs/
output_format = png
output_quality = 95

# Performance Settings
always_low_vram = false
always_offload_from_vram = false

Model Path Configuration

Customize where Fooocus looks for models:

• base_model_path - SDXL base models
• lora_path - LoRA files
• vae_path - VAE models
• upscale_path - Upscaling models
• inpaint_path - Inpainting models

Default Parameters

Set default values that will be used when Fooocus starts:

Generation Parameters

Quality Settings

• default_quality - Quality preset (speed, quality, extreme)
• default_performance - Performance mode
• enable_refiner - Enable refiner by default
• refiner_switch - When to apply refiner

Style Presets

Create and customize style presets for consistent image generation:

Built-in Styles

Fooocus includes several built-in styles:

• Auto: Automatic style selection
• Cinematic: Movie-like quality
• Photographic: Realistic photography
• Anime: Anime/manga style
• Digital Art: Digital artwork style
• Neon Punk: Cyberpunk aesthetic
• Enhance: Quality enhancement

Creating Custom Styles

Define custom styles in your configuration:

# Custom Style Definition
style_custom_name = {
    "prompt": "your style prompt here",
    "negative": "unwanted elements",
    "steps": 30,
    "cfg_scale": 7.0
}

Style Files

Styles can also be defined in separate files in the styles/ directory for better organization.

LoRA Presets

Create LoRA presets for quick application of common LoRA combinations:

Defining LoRA Presets

# LoRA Preset
lora_preset_anime = {
    "loras": [
        {"name": "anime_style", "strength": 1.0},
        {"name": "character_lora", "strength": 0.8}
    ]
}

Using Presets

Apply presets via UI dropdown or command line: --preset anime

Default Prompts

Set default positive and negative prompts that will be used automatically:

Default Positive Prompt

default_positive_prompt = "high quality, detailed, professional"

Default Negative Prompt

default_negative_prompt = "blurry, low quality, artifacts, watermark"

Prompt Templates

Create prompt templates for common use cases:

template_photography = "photorealistic, {subject}, professional photography, 8k, sharp focus"
template_anime = "anime style, {subject}, high quality, detailed"

Output Configuration

Customize how and where Fooocus saves generated images:

Output Settings

• output_path - Directory for generated images
• output_format - Image format (png, jpg, webp)
• output_quality - JPEG quality (1-100)
• save_metadata - Embed generation parameters
• filename_template - Custom filename pattern

Filename Templates

Customize output filenames:

filename_template = "{seed}_{steps}_{cfg_scale}"

Available variables: {seed}, {steps}, {cfg_scale}, {timestamp}

Performance Customization

Optimize Fooocus performance for your hardware:

Memory Settings

• always_low_vram - Always use low VRAM mode
• always_offload_from_vram - Offload models when idle
• max_vram_usage - Maximum VRAM to use
• cpu_offload - Offload to CPU when possible

Speed Settings

• async_cuda_alloc - Async CUDA memory allocation
• attention_optimization - Optimize attention mechanism
• enable_xformers - Use xFormers for faster attention

UI Customization

Customize the user interface appearance and behavior:

Theme Settings

• theme - UI theme (light, dark, auto)
• language - UI language code
• font_size - Base font size

Layout Settings

• show_advanced_options - Show advanced UI options
• compact_mode - Use compact UI layout
• sidebar_position - Sidebar placement

Wildcard Customization

Organize and customize wildcard files:

Wildcard Directory Structure

wildcards/
├── colors/
│   ├── basic.txt
│   └── extended.txt
├── styles/
│   ├── art.txt
│   └── photography.txt
└── subjects/
    ├── animals.txt
    └── objects.txt

Creating Wildcard Files

Create text files with one option per line:

# wildcards/colors/basic.txt
red
blue
green
yellow
orange
purple

Use in prompts: __colors/basic__ flower

Command-Line Presets

Create batch files or scripts for common configurations:

Windows Batch Files

@echo off
REM run_anime.bat
python entry_with_update.py --preset anime --always-low-vram

Linux/macOS Scripts

#!/bin/bash
# run_realistic.sh
python entry_with_update.py --preset realistic --port 7860

Advanced Customization

For advanced users, Fooocus supports deeper customization:

Custom Models

• Add custom SDXL models to checkpoints directory
• Configure model-specific settings
• Create model presets

API Customization

• Custom API endpoints
• Webhook configurations
• Integration settings

Workflow Automation

• Batch processing scripts
• Automated generation pipelines
• Integration with other tools

Configuration Examples

Example configurations for common use cases:

High Quality Configuration

default_steps = 50
default_cfg_scale = 7.0
default_quality = extreme
enable_refiner = true
refiner_switch = 0.8

Fast Generation Configuration

default_steps = 20
default_cfg_scale = 5.0
default_quality = speed
enable_refiner = false

Low VRAM Configuration

always_low_vram = true
always_offload_from_vram = true
max_vram_usage = 4096
cpu_offload = true

Best Practices

Related Resources