Voice cloning · API

Your voice.
A new possibility.

Record a 10-30 second clip and the speaker's spoken consent, POST them to create a voice, then speak in it on Simba 3.5 by its voice_id.

Voice cloning is on paid plans, Starter and up. Your account and API key are free.

Want to listen to your documents in your own voice? That is the Speechify app. It lives at speechify.com

Create a voiceConsent required
bash
curl -X POST https://api.speechify.ai/v1/voices \
  -H "Authorization: Bearer $SPEECHIFY_API_KEY" \
  -H "Speechify-Version: 2026-09-13" \
  -F name="Narrator" \
  -F gender="female" \
  -F sample=@sample.wav \
  -F consent_challenge_id="$CONSENT_CHALLENGE_ID" \
  -F consent_recording=@consent.wav

First create a consent challenge with full_name, then record the speaker reading its phrase. Set CONSENT_CHALLENGE_ID to the returned ID and save the recording as consent.wav before submitting the sample.

  • 10-30s voice sample
  • POST /v1/voices returns a voice_id
  • 48 languages on Simba 3.5, paid plans
  • Zero-shot and fine-tuned
How it works

From a consented sample to a voice_id.

No separate service to integrate. Cloning is an endpoint beside synthesis on the same API, and the voice it creates speaks on Simba 3.5 like any catalog voice.

  1. Record a sample

    10-30 seconds of speech

    Prepare 10-30 seconds of clean speech from the person whose voice you are authorized to clone, in the language you will generate.

  2. Verify consent

    POST /v1/voices/consent-challenges

    Send the speaker's full name as the JSON field full_name. You get back a challenge id and a phrase; record the same speaker reading the phrase.

  3. Create the voice

    POST /v1/voices

    Submit the sample, the consent recording and the challenge ID as multipart form data. The response is your new voice, with its id.

  4. Speak with it

    model: simba-3.5

    Pass that id as voice_id to POST /v1/audio/speech or /v1/audio/stream with simba-3.5, in any of its 48 languages.

Then it is a voice like any other.

Set VOICE_ID to the id that POST /v1/voices returned. Record in the language you will generate: a cloned voice speaks other languages with the accent of its recording.

bash
curl -X POST https://api.speechify.ai/v1/audio/speech \
  -H "Authorization: Bearer $SPEECHIFY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "input": "Every chapter, in my own voice.",
    "voice_id": "VOICE_ID",
    "model": "simba-3.5",
    "language": "en-US",
    "audio_format": "mp3"
  }'
Voice cloning · Listening samples

Hear a clone before you write a line.

Listen to the original speaker and a prepared cloned-voice sample generated with Simba 3.5. Two recordings, different scripts, the same voice identity.

Simba 3.5 · Voice cloning

A new script.
Your voice.

Start with a consented reference recording. Create a reusable voice for the words that come next.

Explore voice cloning

01 / Reference

Original speaker

Ready to listen0:00

02 / Clone

Cloned voice

Ready to listen0:00

Two different recordings. Generated sample: Simba 3.5. Only clone a voice you have permission to use.

Prepared Speechify samples. Choose a sample, then press play.Try your own text
Quality tiers

Two tiers, one API.

Start self-serve, or work with our team on a fine-tuned voice.

Instant · zero-shot

Self-serve cloning

Clone from a 10-30 second sample and a verified consent recording via the API or Console. No fine-tuning run required.

Instant cloning
Fine-tuned

Professional cloning

Fine-tune on hours of a speaker's audio. Arranged with our team for signature narrators and brand voices.

Arrange professional cloning

On a paid plan, a cloned voice speaks all 48 languages of Simba 3.5, under one model id. Record in the language you will generate. A cloned voice speaks other languages with the accent of its recording. The older Simba 3.2 (English) and Simba 3.0 (6 languages) take cloned voices too; check each voice's supported models before synthesis.

Pricing

Part of Build, on one bill.

Voice cloning is included on paid plans, with no separate per-clone charge. Synthesis bills at the model's per-character rate on your plan, exactly like a catalog voice. Free accounts do not include cloning.

per 1M characters · Simba 3.5 on Starter
List price $20 launch price $10
cloning included
Starter
voice sample guidance
10-30s
languages on Simba 3.5
48

Simba 3.5 lists at $20/$16/$12 per 1M characters on Starter/Pro/Scale, billed at $10/$8/$6 until 21 Nov 2026. Simba 3.0 and 3.2 bill $10/$8/$6 per 1M characters.

Use cases

Where a cloned voice earns its keep.

Accessibility

Assistive technology that speaks in a person's own voice instead of a generic one: clone from a short sample with consent and use the voice ID in read-aloud and communication tools, or preserve a voice before it is lost.

Audiobooks

An author or narrator reads a whole book in their own voice, generated from the manuscript. One voice ID keeps every chapter consistent, and the same clone reads the translated editions.

Creator tools

A cloning feature inside your own product: your users clone their voice and generate content in it. The API provides the endpoints and enforces consent; you provide the flow and manage the voice IDs.

Podcasts

One signature host voice across every episode, produced from a script, with cross-language editions in the same voice rather than a different presenter per market.

Game characters

A signature character keeps one consistent voice across every line and every update, with new lines synthesized on demand from a consented clone of the actor.

FAQ

Common questions.

What is voice cloning?
It creates a synthetic version of a specific voice from an audio sample, then synthesizes any text in that voice. On the Build API you clone from a short clean sample, with the speaker's consent, and use it by voice ID with simba-3.5 on the normal speech endpoints, in any of its 48 languages. It is a Build feature that shares the Build API and pricing.
Do I need consent to clone a voice?
Yes. Create a consent challenge and record the speaker reading its phrase. Submit that recording with the challenge ID and voice sample. The recording is checked against the phrase and speaker, then retained as the consent record.
What languages do cloned voices support?
On a paid plan, a cloned voice speaks all 48 languages of Simba 3.5 under one model id, simba-3.5. Record in the language you will generate. A cloned voice speaks other languages with the accent of its recording. The older Simba 3.2 (English) and Simba 3.0 (6 languages) take cloned voices too.
Instant or professional cloning: which one?
Instant cloning is zero-shot: a 10-30 second voice sample and a verified consent recording create a reusable voice without a fine-tuning run. Professional cloning fine-tunes on hours of a speaker's audio and is arranged with our team for signature narrators and brand voices.
What does voice cloning cost?
Nothing separate. Cloning is a Build capability included on Starter, Pro, Scale and Enterprise (a Free-tier attempt returns a 402 with the code voice_cloning_not_included), and speaking with a cloned voice is billed per character at the model's rate, exactly like a catalog voice. Simba 3.5 lists at $20/$16/$12 per 1M characters on Starter/Pro/Scale, billed at $10/$8/$6 until 21 Nov 2026. Simba 3.0 and 3.2 bill $10/$8/$6 per 1M characters. There is no per-clone price and no premium for a cloned voice.
How do I manage cloned voices?
Retrieve a voice with GET /v1/voices/{voice_id}, download its sample as WAV, and remove it with DELETE /v1/voices/{voice_id}. Cloned voices list before shared voices.

Build with a verified voice.

Create an account and API key, choose a paid plan, then record a sample and verify the speaker's consent.

Voice cloning is on paid plans, Starter and up. Your account and API key are free.

Privacy preferences

Choose what we may store on this device. You can change this at any time from the footer.

Strictly necessary

Sign-in, security, load balancing, and remembering your privacy choices. These cannot be switched off.

Always on

Analytics

How the site is used in aggregate - which pages get read, where people get stuck - so we can improve it.

Marketing

Measures which campaigns bring people here, and lets us show relevant ads on other platforms.