Piper TTS Web Interface

Note

This project has been retired. Basic TTS is no longer running as a live service, and the code is no longer actively maintained. The repository is kept available as a reference for anyone who wants to learn from it or build their own text-to-speech system. Thank you to everyone who used it! 🎙️

This is a simple web interface for Piper Text-to-Speech that provides a user-friendly way to convert text to speech using various voice models.

This repo was almost entirely vibe-coded in Cursor.

Prerequisites

1. Python 3.8 or higher
2. espeak-ng (required for Piper)
3. Git LFS (for managing voice model files)
4. Voice models in ONNX + JSON format

Installation

Local Development

1. Install system dependencies:

   For macOS:

   # Install espeak-ng, git-lfs and build dependencies
   brew install espeak-ng cmake ninja git-lfs
   
   # Initialize Git LFS
   git lfs install
   
   # Install Piper (much simpler now!)
   pip install piper-tts

   For Linux:

   # Install espeak-ng and git-lfs
   sudo apt-get update
   sudo apt-get install espeak-ng git-lfs
   
   # Initialize Git LFS
   git lfs install
   
   # Install Piper (much simpler now!)
   pip install piper-tts

   For Windows:

   # Install Git LFS using the installer from https://git-lfs.com
   # Or using Chocolatey:
   choco install git-lfs
   
   # Initialize Git LFS
   git lfs install
   
   # Install Piper
   pip install piper-tts

2. Install the web interface package:

   pip install -e .

3. Download voice files from Piper Samples (you can listen to samples and download the ones you like)

   • Place the downloaded .onnx files and their corresponding .json files in the models directory

Cloud Deployment (disco.cloud)

The application is deployed to the cloud using disco.cloud with automatic Docker containerization.

1. Configure disco.json for the service:

   {
     "version": "1.0",
     "services": {
       "web": {
         "port": 8000
       }
     }
   }

2. Set environment variables via the dashboard.disco.cloud UI:

   • FIREBASE_API_KEY
   • FIREBASE_AUTH_DOMAIN
   • FIREBASE_PROJECT_ID
   • FIREBASE_STORAGE_BUCKET
   • FIREBASE_MESSAGING_SENDER_ID
   • FIREBASE_APP_ID
   • FIREBASE_MEASUREMENT_ID
   • FIREBASE_SERVICE_ACCOUNT_JSON
   • REVENUECAT_API_KEY

3. Deploy using disco:

   disco deploy

Local Docker Deployment

For local Docker deployment:

1. Build the Docker image:

   docker build -t piper-tts-web .

2. Run the container:

   docker run -p 8000:8000 -v $(pwd)/models:/app/models piper-tts-web

Usage

Cloud Deployment

The application is deployed at BasicTTS.com

Local Development

1. Start the development server:

   python -m piper_tts_web.server

   The server will start on http://localhost:8000

2. Open your web browser and navigate to http://localhost:8000

3. Select a voice from the dropdown menu

4. Enter the text you want to convert to speech

5. Click "Convert to Speech" and wait for the audio to be generated

6. Use the audio player to:

   • Play/pause the generated audio
   • Adjust the playback speed
   • Download the audio file

Development

Setting up a development environment

1. Create a virtual environment (optional but recommended):

   python -m venv venv
   source venv/bin/activate  # On macOS/Linux

2. Install development dependencies:

   pip install -e ".[dev]"

3. Run tests:

   pytest

4. Format code:

   black .

Troubleshooting

1. If you get a "piper command not found" error:

   • Make sure espeak-ng is installed
   • Verify that Piper is installed correctly
   • Check that the package is in your Python path
   • Try running which piper to verify the installation
   • Make sure your PATH includes the directory where piper is installed

2. If no voices appear in the dropdown:

   • Check that you have .onnx files in the models directory
   • Make sure each .onnx file has a corresponding .json file
   • Verify that the files are properly named (e.g., en_US-amy-medium.onnx)

3. If the conversion fails:

   • Check the server logs for detailed error messages
   • Verify that the selected voice model exists
   • Make sure the text input is not empty

Notes

• The server expects voice model files to be in ONNX format
• Temporary audio files are automatically cleaned up after processing
• The web interface supports various playback speeds (0.25x to 2x)
• Processing can be done locally or in the cloud
• Docker deployment provides consistent environments
• Cloud deployment via disco.cloud for production
• Firebase authentication integrated for user management
• RevenueCat integration for subscription management