Introducing Song Creator Pro — create music with AI, locally on your device. Try it now →
Voice Creator Pro logoVoice Creator Pro
Try For Free

Voice Cloning API

Voice cloning API: clone a voice from a few seconds of audio

Save a voice once from a few seconds of audio, then send its ID with any text to hear it speak.

Available now as a local API in the desktop app. Cloud API coming soon.

Join the cloud API waitlistGet the desktop app
POST/api/v1/generation/clone
curl -X POST "http://127.0.0.1:8100/api/v1/generation/clone?sync=true" \
  -H "Content-Type: application/json" \
  -d '{
    "voice_id": "VOICE_ID",
    "target_text": "Thanks for watching. See you next week.",
    "language": "English"
  }'

What developers build with it

Cloning is zero-shot: no training step, no waiting. A clean clip of 3 to 10 seconds gives the best match.

  • A creator's own voice reading new scripts, without recording each one
  • Consistent character voices across every episode of a series
  • Localized lines in the original speaker's voice, in any of 600+ languages
See the product page
Response
{
  "success": true,
  "status": "Generated successfully",
  "audio_base64": "UklGRiQAAABXQVZFZm10..."
}

Need voice cloning as a cloud endpoint for your users?

Join the cloud API waitlist

Endpoints

Voice cloning endpoints

MethodPathWhat it does
POST/api/v1/voicesSave a reference clip and transcript as a voice
POST/api/v1/generation/cloneGenerate speech in a cloned voice
POST/api/v1/generation/clone/streamStream cloned speech as it generates
GET/api/v1/voicesList saved voices

Quickstart

Make your first request in three steps

Install the app, turn on the API and send your first request.

API reference

Full API reference

Every endpoint, with its parameters and responses.

Cloud API

Join the cloud API waitlist

We're building a cloud API. Pick the features you need and roughly how much you'd use, and your answers decide which ones come first.

Which features do you need?

We'll only email you about the API. Unsubscribe anytime.

FAQ

Common Questions

3 to 10 seconds of clear speech gives the best clone. Longer clips don't improve it. Include the transcript of the clip as ref_text for the closest match.

Yes. Save the clip with POST /api/v1/voices, then pass the returned voice_id to /api/v1/generation/clone. The saved audio is used automatically on every request.

Yes, with the Qwen3-TTS model: set emotion and intensity on the request to choose from 13 emotions at 5 intensities. With OmniVoice, emotion comes from the reference clip instead.

Yes. POST /api/v1/generation/clone/stream returns audio as it is generated, so playback can start before the full line is finished.

Explore

Related APIs

Text to speech

Turn text into speech with preset speakers in 10 languages.

Voice design

Describe a voice in words and generate speech in it.

Speech to text

Transcribe audio with word timestamps and speaker labels.

Dubbing

Dub videos into 21 languages, keeping each speaker's voice.

Subtitles

Generate SRT caption files with word-level timing.

Clips

Cut long videos into short, captioned clips.

AI Cut

Remove silences, filler words and bad takes, then render.

Avatars

Make a lip-synced talking-head video from one photo.

Voice changer

Re-voice a recording while keeping its timing and emotion.

Audiobooks

Narrate a manuscript with a voice for every character.