This repository is a fork of the original DMOSpeech2 repository. The original README has been renamed to original-README.md.

• uv - Python package installer and virtual environment manager

1. Activate the virtual environment:

   source setup-source-me.sh

2. Run the setup script:

   ./scripts/setup.sh

   Note: This downloads large model files (~500MB each) to the ckpts/ directory from HuggingFace.

Try the original DMOSpeech2 online without any setup:

• HuggingFace Spaces (Original Repository): https://huggingface.co/spaces/yl4579/DMOSpeech2-demo

For quick testing without local setup, use the Google Colab notebook:

• Open DOMSpeech2_gradio_colab_GPU.ipynb in Google Colab
• Run all cells to set up environment and launch Gradio interface
• Provides free GPU access for faster inference

For single-machine development and testing. Services bind to 127.0.0.1 (localhost only) for security.

python scripts/local-fastapi.py

• Access API at: http://127.0.0.1:8000
• API documentation: http://127.0.0.1:8000/docs

python scripts/local-gradio.py

• Access UI at: http://127.0.0.1:7860

./scripts/jupyter-lab-local.sh

• Access Jupyter at: http://127.0.0.1:8888

Three notebook demos are available:

1. src/serveDMO.ipynb - FastAPI demo

   • Run the cell to start FastAPI server on port 8000

2. src/gradio-test.ipynb - Gradio UI demo

   • Run the cell to start Gradio interface on port 7860

3. DOMSpeech2_gradio_colab_GPU.ipynb - Google Colab demo with GPU support

   • Run DMOSpeech2 in Google Colab with free GPU access
   • Includes all necessary setup and Gradio interface

To access local services from a remote machine, use SSH port forwarding:

# From your remote machine to access local services
ssh -L 7860:localhost:7860 -L 8000:localhost:8000 user@hostname

# Then access in your remote browser:
# - Gradio UI: http://localhost:7860  
# - FastAPI docs: http://localhost:8000/docs

This enables microphone access and full UI functionality from remote browsers while maintaining security.

⚠️ Security Warning: These scripts expose services to your local network. Only use on trusted networks behind firewalls.

python scripts/remote-fastapi.py

• Access from any device on your network: http://YOUR_IP:8000

python scripts/remote-gradio.py

• Access from any device on your network: http://YOUR_IP:7860

./scripts/jupyter-lab-remote.sh

• Access from any device on your network: http://YOUR_IP:8888

# Initialize voice with reference audio
curl -X POST "http://127.0.0.1:8000/init_voice" \
  -F "audio_file=@reference.wav" \
  -F "reference_text=Your reference text here"

# Generate speech from text
curl -X POST "http://127.0.0.1:8000/generate_audio" \
  -F "target_text=This is the text I want synthesized." \
  --output generated_audio.wav

For network access, replace 127.0.0.1 with your server's IP address.

• ✅ Secure: Services only accessible from the same machine
• ✅ Recommended: For development and testing
• ✅ Safe: No network exposure

• ✅ Secure: Encrypted connection to remote services
• ✅ Flexible: Access remote services as if they were local
• ✅ Best Practice: For remote access to development servers

• ⚠️ Caution Required: Exposes services to local network
• ⚠️ Firewall Needed: Ensure proper network security
• ⚠️ No Authentication: Services have no built-in security
• ⚠️ HTTP Only: No encryption (consider HTTPS for production)

For production use, consider:

• HTTPS/SSL certificates
• Authentication and authorization
• Rate limiting and monitoring
• Reverse proxy (nginx, Apache)
• Network security hardening

1. Port already in use: Change port numbers in scripts if conflicts occur
2. Permission denied: Ensure scripts are executable (chmod +x scripts/*.sh)
3. Module not found: Verify virtual environment is activated
4. CUDA errors: Check GPU availability and PyTorch installation

• Check the original documentation in original-README.md
• Review error logs for specific issues
• Ensure all prerequisites are installed

• Original DMOSpeech2 repository: DMOSpeech2
• Additional codebase references: F5-TTS, DMD2, simple_GRPO

This fork aims to provide enhanced ease-of-use and seamless integration of DMOSpeech2 into broader workflows and user interfaces.