SayServer

Web service for generating TTS (text-to-speech).

This package is built with streaming use cases in mind and makes a few opinionated choices. Every TTS request requires a seed (often a username), which is used to consistently select from a pool of random voices—so each user has their own unique "voice." For non-user messages, you can explicitly choose a specific voice instead.

The server can provide both locally generated TTS and remote content—for example, voices from the public Streamlabs Polly server.

Usage

The server runs using PM2. See the API documentation on how to call the server.

Supported sources

• SpeechSynthesizer for local generation on macOS (os.platform must be "darwin").
• Amazon Polly for remote generation, using the Streamlabs proxy.

External links

• lazypy.ro TTS tester
• macOS say command documentation
  • Techniques for Customizing Synthesized Speech
• wsay, Windows "say"

License

MIT license.