README.md

metadata

title: Stable Audio Open
emoji: 🎵
colorFrom: blue
colorTo: purple
sdk: gradio
sdk_version: 6.2.0
app_file: app.py
pinned: false

🎵 Stable Audio Open

An open-source web interface for generating high-quality audio from text prompts using advanced AI models. Create music, sound effects, ambient sounds, and more with simple text descriptions.

Features

• 🎼 Text-to-Audio Generation: Convert text prompts into audio
• 🎚️ Customizable Duration: Generate audio from 1-30 seconds
• 🎲 Reproducible Results: Use seeds for consistent generation
• 🎧 Real-time Playback: Listen to generated audio instantly
• 📝 Example Prompts: Pre-built examples to get you started

Setup

Hugging Face Authentication (Recommended)

For the best experience and to avoid rate limits, set up Hugging Face authentication:

1. Create a Hugging Face account at huggingface.co
2. Go to Settings > Access Tokens and create a new token
3. Copy env-example.txt to .env and add your token:

   HF_TOKEN=hf_your_token_here
   

4. Install dependencies: pip install -r requirements.txt

Note: Without authentication, you may experience rate limits or reduced access to some models.

Usage

1. Enter a text description of the audio you want to generate
2. Adjust the duration slider (1-30 seconds)
3. Optionally set a random seed for reproducible results
4. Click "Generate Audio" to create your sound

Examples

• "A gentle piano melody playing in a cozy room"
• "Upbeat electronic dance music with synthesizers"
• "Rain falling on a tin roof with distant thunder"
• "Classical violin concerto with orchestra accompaniment"

Technical Details

This application uses:

• AudioLDM2 (cvssp/audioldm2) - Advanced AI model for text-to-audio generation
• Gradio for the web interface
• PyTorch & Diffusers for model inference
• NumPy for audio processing and fallback synthesis
• Automatic fallback to simple synthesis if model is unavailable

Model Information

• Model: cvssp/audioldm2
• First Run: Model will be automatically downloaded (~1-2 GB)
• Device: Automatically uses GPU if available, falls back to CPU
• Caching: Model is cached in memory for faster subsequent generations

Contributing

This is an open-source project. Contributions are welcome! Feel free to:

• Report bugs and issues
• Suggest new features
• Submit pull requests
• Improve documentation

License

This project is open source and available under the MIT License.

---

Built with ❤️ using Hugging Face Spaces and Gradio