@mastra/speech-replicate

Replicate Speech integration for Mastra, providing Text-to-Speech (TTS) capabilities using various AI models hosted on Replicate.

Installation

npm install @mastra/speech-replicate

Configuration

The module requires the following environment variable:

REPLICATE_API_TOKEN=your_api_token

Usage

import { ReplicateTTS } from '@mastra/speech-replicate';

// Initialize with configuration
const tts = new ReplicateTTS({
  model: {
    name: 'default', // Default model
    apiToken: 'your-api-token', // Optional, can use REPLICATE_API_TOKEN env var
  },
});

// List available models
const voices = await tts.voices();

// Generate speech
const result = await tts.generate({
  voice: 'model-id',
  text: 'Hello from Mastra!',
});

// Stream speech
const stream = await tts.stream({
  voice: 'model-id',
  text: 'Hello from Mastra!',
});

Features

• Access to multiple TTS models
• High-quality speech synthesis
• Streaming support
• Model version control
• Various voice options

Model Options

Replicate provides access to various TTS models:

• Coqui TTS
• Bark
• Tortoise
• Facebook MMS
• And more...

Each model has its own unique characteristics and capabilities. View the available models in the voices.ts file or Replicate's model collection.

Additional Features

• Model version selection
• Custom model parameters
• Multiple output formats
• Batch processing
• Progress tracking