FLUX LoRA DLC 🥳

Demo.mp4

This repository hosts a Gradio web interface designed to explore the capabilities of the FLUX.1-dev diffusion model combined with an extensive collection of community-created LoRAs (Low-Rank Adaptations). It provides an easy way to generate images in various styles defined by these LoRAs, supporting both Text-to-Image and Image-to-Image generation.

The "DLC" (Downloadable Content) in the name playfully refers to the massive, ever-growing library of LoRAs integrated into the app, allowing users to instantly "download" and apply different artistic styles.

Features

• FLUX.1-dev Integration: Leverages the powerful black-forest-labs/FLUX.1-dev model.
• Massive LoRA Library: Includes over 200+ pre-configured LoRAs sourced from the Hugging Face Hub community.
• LoRA Gallery: Easily browse and select LoRAs with preview images and titles.
• Custom LoRA Support: Load any compatible FLUX LoRA directly from a Hugging Face repository link.
• Text-to-Image Generation: Create images from text prompts using selected LoRAs.
• Image-to-Image Generation: Modify existing images based on text prompts and LoRA styles.
• Real-time Preview: Utilizes the TAEF1 tiny autoencoder for a fast preview during the generation process (for Text-to-Image).
• User-Friendly Interface: Built with Gradio for simple interaction.
• Adjustable Parameters: Control steps, CFG scale, seed, dimensions, LoRA scale, and image strength (for I2I).
• Performance Monitoring: Includes basic timing for LoRA loading and image generation.

Setup and Installation

Prerequisites:

• Python 3.8+
• pip and git
• A CUDA-enabled GPU with sufficient VRAM (at least 16GB recommended, more might be needed depending on resolution) and installed NVIDIA drivers.

Installation Steps:

1. Clone the repository:

   git clone https://github.com/PRITHIVSAKTHIUR/FLUX-LoRA-DLC.git
   cd FLUX-LoRA-DLC

2. Create a virtual environment (Recommended):

   python -m venv venv
   source venv/bin/activate  # On Windows use `venv\Scripts\activate`

3. Install dependencies:

   pip install -r requirements.txt

   (See requirements.txt section below if you need to create it)

4. Hugging Face Login (Optional but Recommended): To ensure seamless access to models and LoRAs from the Hugging Face Hub, log in using your token:

   huggingface-cli login

   Alternatively, you can uncomment and add your token directly in the script (less secure).

5. Run the Gradio App:

   python app.py

   (Assuming your main script is named app.py. Adjust if necessary.)

   The interface should now be accessible via a local URL (usually http://127.0.0.1:7860).

requirements.txt

If a requirements.txt file is not included, create one with the following content (versions might need adjustment based on compatibility):

torch --index-url https://download.pytorch.org/whl/cu121 # Or adjust cuXXX for your CUDA version
diffusers>=0.29.0 # Check for latest compatible version
transformers>=4.40.0 # Check for latest compatible version
gradio>=4.0.0
huggingface_hub>=0.20.0
Pillow
numpy
accelerate
safetensors
invisible_watermark
omegaconf

Usage Guide

1. Launch the app: Run python app.py.
2. Select a LoRA:
   • Click on a LoRA card in the gallery. The selected LoRA's Hugging Face repo link will appear above the gallery.
   • OR enter a Hugging Face repository link (e.g., username/my-flux-lora) into the "Enter Custom LoRA" textbox and press Enter. A card will appear confirming the loaded custom LoRA and its trigger word (if found).
3. Enter Prompt: Type your desired image description in the "Prompt" box.
   • Important: Check the selected LoRA information or the custom LoRA card. If a "trigger word" is specified, include it in your prompt for the LoRA style to activate correctly (e.g., prompt text, trigger_word).
4. Configure Settings (Optional):
   • Image-to-Image: Upload an Input image under "Advanced Settings" and adjust the Denoise Strength (lower values preserve more of the original image).
   • Text-to-Image: Adjust Steps, CFG Scale, Width, Height, LoRA Scale.
   • Seed: Use the Seed slider or check Randomize seed for unique results each time.
5. Generate: Click the "Generate" button.
6. View Result:
   • For Text-to-Image, a real-time preview using TAEF1 will update in the result area, followed by the final high-quality image decoded with the full VAE. A progress bar indicates the steps.
   • For Image-to-Image, the final image will appear after processing.

Key Components

• Base Model: black-forest-labs/FLUX.1-dev
• Preview Autoencoder: madebyollin/taef1
• High-Quality Autoencoder: VAE from black-forest-labs/FLUX.1-dev
• Core Library: diffusers by Hugging Face
• Interface: Gradio

LoRA Credits

This application heavily relies on the amazing work of the AI art community who train and share these LoRAs. Huge thanks to all the creators whose LoRAs are featured in the default list and available on the Hugging Face Hub!

You can find more FLUX LoRAs here: https://huggingface.co/models?other=base_model:adapter:black-forest-labs/FLUX.1-dev

Contributing

Contributions are welcome! Feel free to open an issue or submit a pull request if you have suggestions for improvements, bug fixes, or want to add more default LoRAs (please ensure they are compatible and appropriately licensed).

License

This project is licensed under the Apache License 2.0. See the LICENSE file for details. (You'll need to add an Apache 2.0 LICENSE file to your repo).