AI Voice Cloning Text to Speech — Expressive English Voices

Clone a voice from one short recording and make it read any English script — with natural expression, including laughs and sighs. This is the quality-first option when the voice has to sound convincingly human.

Good for: branded narration in your own voice, character voices for stories and games, podcast and YouTube voiceovers, ads, audiobook samples, dubbing tests.

Quick start

{
  "text": "Oh, that is hilarious! [laugh] Anyway, how are you today?",
  "ref_audio_url": "https://example.com/my-voice.wav",
  "format": "mp3"
}

No reference clip? Leave ref_audio_url empty and a built-in voice is used.

What you get

• Voice cloning from a clean 5–15 second English recording
• Expressive delivery with inline cues such as [laugh], [chuckle], [cough]
• Built-in default voice when you don't supply a reference
• Long text supported — split and joined with smooth transitions
• Speed control, silence trimming, MP3 / WAV / Opus output
• One dataset row per run with the audio URL and timings

Input

Field	Default	What it does
text	required	The script to read
ref_audio_url	—	Link to the voice you want to clone
speed	1.0	Playback speed
dtype	fp32	Use a quantized option to trade a little quality for speed
format	wav	wav, mp3 or opus

Pricing

No rental fee — you only pay Apify compute. This Actor prioritises quality, so a long script takes longer than the fast voice Actors; for bulk jobs use the Fast Text to Speech Actor instead.

Notes

• English only. For Vietnamese use the Vietnamese Text to Speech Actor.
• Only clone voices you own or have permission to use.

For AI agents (MCP-ready)

This actor is built to be called by AI agents. It works out of the box with the Apify MCP Server — add it to Claude Desktop, Cursor or any MCP client: the agent can clone an English voice on its own.

{
  "mcpServers": {
    "chatterbox-tts": {
      "url": "https://mcp.apify.com/?actors=andrew_babo/chatterbox-tts",
      "headers": { "Authorization": "Bearer <YOUR_APIFY_TOKEN>" }
    }
  }
}

Agent skill (paste into your agent's instructions)

Use the "chatterbox-tts" tool when ENGLISH speech must sound convincingly
human, or must be read in a specific cloned voice, including expressive
cues like [laugh] or [sigh]. It is the quality-first option and slower than
"kokoro-tts" / "piper-tts" — use those for bulk or draft work.

HOW TO CALL
- { "text": "Oh, that is hilarious! [laugh] Anyway...",
    "ref_audio_url": "https://.../my-voice.wav", "format": "mp3" }
- No reference clip: omit "ref_audio_url" and a built-in voice is used.
- "dtype": a quantized option trades a little quality for speed.

OUTPUT CONTRACT
- One row per run: audio_url, duration, processing time.
- English only — route Vietnamese to "vieneu-tts", other languages to
  "qwen3-tts" or "kokoro-tts".
- errorCode / noResults means empty input.
- Only clone voices the user owns or has permission to use.