Hume TTS

Emotionally expressive text-to-speech that can convey tone, emotion, and personality.

Property	Value
Model ID	`hume/tts`
Provider	Hume
Type	Text-to-speech

Basic Usage

import { compose, generateAudio, audioModel } from "@synthome/sdk";

const execution = await compose(
  generateAudio({
    model: audioModel("hume/tts", "hume"),
    text: "I'm so excited to share this news with you!",
  }),
).execute();

Emotion Control

Hume TTS understands emotional context and delivers appropriate prosody:

// Excited delivery
generateAudio({
  model: audioModel("hume/tts", "hume"),
  text: "We just hit a million subscribers!",
  description: "Excited, energetic, celebratory",
});

// Calm delivery
generateAudio({
  model: audioModel("hume/tts", "hume"),
  text: "Take a deep breath and relax.",
  description: "Calm, soothing, gentle",
});

// Serious delivery
generateAudio({
  model: audioModel("hume/tts", "hume"),
  text: "This is an important safety announcement.",
  description: "Serious, authoritative, clear",
});

Options

Option	Type	Default	Description
`text`	`string`	required	Text to synthesize
`voice`	`object`	-	Voice specification
`description`	`string`	-	Emotional/style description
`speed`	`number`	-	Speed multiplier (0.5-2.0)
`trailingSilence`	`number`	-	Silence after speech (0-5s)
`format`	`"mp3" \| "wav" \| "pcm"`	`mp3`	Output format
`version`	`"1" \| "2"`	auto	Model version
`numGenerations`	`number`	-	Batch generation (1-5)

Voice Selection

By Voice ID

generateAudio({
  model: audioModel("hume/tts", "hume"),
  text: "Using a specific voice.",
  voice: {
    id: "your-voice-id-here",
    provider: "HUME_AI",
  },
});

By Voice Name

generateAudio({
  model: audioModel("hume/tts", "hume"),
  text: "Using a named voice.",
  voice: {
    name: "Friendly Assistant",
    provider: "HUME_AI",
  },
});

Custom Voices

generateAudio({
  model: audioModel("hume/tts", "hume"),
  text: "Using my custom cloned voice.",
  voice: {
    id: "custom-voice-id",
    provider: "CUSTOM_VOICE",
  },
});

Description as Acting Direction

When using with a voice, the description acts as acting direction:

generateAudio({
  model: audioModel("hume/tts", "hume"),
  text: "I can't believe you did that.",
  voice: { name: "Default Voice" },
  description: "Playfully surprised, with a hint of disbelief", // Acting direction
});

Keep descriptions under 100 characters for best results.

Voice Generation from Description

Without a voice, the description generates a new voice:

generateAudio({
  model: audioModel("hume/tts", "hume"),
  text: "Welcome to our meditation session.",
  description: "Soft female voice, warm and calming, slight British accent",
});

Speed Control

// Slower for emphasis
generateAudio({
  model: audioModel("hume/tts", "hume"),
  text: "Let me explain this carefully.",
  speed: 0.8,
});

// Faster for energy
generateAudio({
  model: audioModel("hume/tts", "hume"),
  text: "Quick update for everyone!",
  speed: 1.3,
});

Recommended range: 0.75-1.5 for stability.

Trailing Silence

Add silence after the speech:

generateAudio({
  model: audioModel("hume/tts", "hume"),
  text: "End of chapter one.",
  trailingSilence: 2, // 2 seconds of silence
});

Batch Generation

Generate multiple variations:

generateAudio({
  model: audioModel("hume/tts", "hume"),
  text: "Choose the best take.",
  numGenerations: 3, // Get 3 different renditions
});

Examples

Podcast Intro

generateAudio({
  model: audioModel("hume/tts", "hume"),
  text: "Welcome back to Tech Talk, the podcast where we explore the future of technology.",
  description: "Warm, enthusiastic, professional podcast host",
});

Meditation Guide

generateAudio({
  model: audioModel("hume/tts", "hume"),
  text: "Close your eyes. Take a deep breath in... and slowly release.",
  description: "Calm, gentle, soothing, slow pace",
  speed: 0.85,
  trailingSilence: 3,
});

Character Voice

generateAudio({
  model: audioModel("hume/tts", "hume"),
  text: "Greetings, traveler! What brings you to our village?",
  description: "Friendly elderly wizard, wise and welcoming, slight rasp",
});

Hume TTS

On this page