Voice Cloning API

Voice cloning creates a synthetic version of a specific voice from a short sample, then synthesizes any text in that voice. It is part of the Build API: you create a voice, get an ID, and use it on the same speech endpoints as any catalog voice.

Cloning happens in two legs. The consent leg proves the speaker agreed: a challenge phrase, read aloud, recorded. The cloning leg, this page, turns the sample plus that proof into a voice. The Consent page covers the first leg in full - the challenge lifecycle, the recording rules, and every verification refusal - and the sequence diagram shows how the two calls fit together.

$
export SPEECHIFY_API_KEY="your-api-key-here"

1
curl -X POST https://api.speechify.ai/v1/voices/consent-challenges \
2
     -H "Authorization: Bearer <token>" \
3
     -H "Content-Type: application/json" \
4
     -d '{
5
  "full_name": "Jane Doe"
6
}'

1
curl -X POST https://api.speechify.ai/v1/voices \
2
     -H "Authorization: Bearer <token>" \
3
     -H "Content-Type: multipart/form-data" \
4
     -F avatar=@<file1> \
5
     -F consent_challenge_id="string" \
6
     -F consent_recording=@string \
7
     -F gender="male" \
8
     -F name="string" \
9
     -F sample=@string

1
curl -X POST https://api.speechify.ai/v1/audio/speech \
2
     -H "Authorization: Bearer <token>" \
3
     -H "Content-Type: application/json" \
4
     -d '{
5
  "input": "Hello! This is the Speechify text-to-speech API.",
6
  "voice_id": "geffen_32",
7
  "audio_format": "mp3",
8
  "model": "simba-3.2"
9
}'

Required fields

When verification fails

A create refused on consent grounds returns a consent_* error code, and three of them share 422, so branch on the code rather than the status. Every code, what it means, and the fix is in When verification fails on the Consent page. One rule to build around: a refused create still spends the challenge, so every retry starts with a new challenge and a new phrase.

Versioning

This flow ships on Speechify-Version: 2026-09-13; it is callable now by pinning that version, and it becomes the default for new workspaces on that date. Workspaces pinned to earlier versions get the previous shape, where POST /v1/voices instead takes a consent form field (a JSON string with fullName and email) and no challenge. That flow is deprecated as of 13 August 2026 and will be switched off on a date announced in the changelog, deliberately sooner than the standard 12-month sunset because the unverified flow is a safety liability rather than just an old shape. A pinned default never moves on its own, so migrating means re-pinning 2026-09-13.

Model support

Cloned voices run self-serve on simba-3.0, simba-english and simba-multilingual. Use simba-multilingual to speak a cloned voice across 30+ languages from one voice ID.

Next steps