Async Audio TTS

Minimal TypeScript interface for Text-to-Speech engines that output audio as an asynchronous stream. Built to interoperate with @silyze/async-audio-stream.

Install

npm install @silyze/async-audio-tts

This package depends on @silyze/async-audio-stream and re-exports its types (e.g., AudioOutputStream, AudioFormat).

Quick Start

@silyze/async-audio-tts exports a single interface describing a TTS engine that produces audio you can read asynchronously.

import TextToSpeachModel from "@silyze/async-audio-tts";
import { AudioFormat } from "@silyze/async-audio-stream";

class YourEngine implements TextToSpeachModel {
  // AudioOutputStream requirement
  get format(): AudioFormat {
    // Return the format of emitted audio (e.g., PCM 16kHz mono)
    throw new Error("Not implemented");
  }

  // TTS-specific lifecycle
  get ready(): Promise<void> {
    // Resolve when the engine is initialized and ready
    return Promise.resolve();
  }

  async speak(text: string): Promise<void> {
    // Enqueue synthesized audio for the provided text into the output stream
    throw new Error("Not implemented");
  }

  async close(): Promise<void> {
    // Release internal resources and stop emitting audio
    throw new Error("Not implemented");
  }

  // Provide async iteration of audio chunks (Buffer)
  async *[Symbol.asyncIterator](): AsyncIterator<Buffer> {
    // Yield audio buffers as they become available
    // yield chunk
  }
}

Consume audio from any TextToSpeachModel implementation:

const tts = new YourEngine();
await tts.ready;
await tts.speak("Hello world");

for await (const chunk of tts) {
  // Handle PCM/encoded audio Buffer chunks (e.g., write to a file or stream to a player)
}

await tts.close();

API

• Default export TextToSpeachModel extends AudioOutputStream and adds:
  • ready: Promise<void> - resolves when the engine is initialized and ready to synthesize.
  • speak(text: string): Promise<void> - synthesizes and enqueues audio for text onto the output stream.
  • close(): Promise<void> - stops synthesis and releases any resources used by the engine.

From @silyze/async-audio-stream:

• AudioOutputStream - an async readable stream of Buffer chunks with a format: AudioFormat getter.
• AudioFormat - describes the encoding of the emitted audio (e.g., PCM sample rate, encoding name).

Type Definition

import { AudioOutputStream } from "@silyze/async-audio-stream";

export default interface TextToSpeachModel extends AudioOutputStream {
  get ready(): Promise<void>;
  speak(text: string): Promise<void>;
  close(): Promise<void>;
}

Notes

• Multiple speak() calls may enqueue additional audio onto the same output stream.
• Consumers should continuously read from the stream to avoid backpressure.
• Call tts.close() when finished to free resources and stop audio production.
• The emitted audio format is discoverable via tts.format.