Voice Cloning
Overview
Clone your own voice, or one you have the speaker’s permission to clone, from a 10-30 second audio sample. The cloned voice captures accent, speaking style, and tone. There’s no limit to the number of voices you can create.
Creating a voice is a two-leg system, because consent is captured as part of it:
- The consent leg. Speechify issues a single-use challenge phrase, the speaker reads it aloud, and you record them doing so. That recording is the consent record. Full detail on the Consent page.
- The cloning leg. You send the voice sample, the recording, and the challenge id to one create call; Speechify verifies the recording against the phrase it issued, and the voice comes back ready to synthesize with. Full detail on the Voice Cloning API page.
Neither leg works alone: a challenge with no create expires unused, and a create with no verified recording is refused. The quickstart walks both legs end to end.
The previous flow, a consent form field carrying the speaker’s fullName and email as a JSON string, is deprecated and will be switched off on a date announced in the changelog. Existing cloned voices are unaffected.
How the flow works
Two things follow from the shape of this. The challenge is single use and short-lived, so create it when your speaker is ready to record rather than at the start of your flow - and if an attempt is refused, that challenge is gone, so the retry begins with a new one. And the consent recording is not a second voice sample: it is the evidence, it must be the same person as the sample, and it is retained with the voice.
The consent leg: challenge lifecycle, recording rules, verification failures, rate limits.
The cloning leg: create the voice, required fields, model support, synthesize with it.
After creation: list, inspect, download the sample, delete.
Both legs end to end, from sample to synthesized speech.
Cloned voices work self-serve on simba-3.0, simba-english and simba-multilingual. simba-3.2 serves cloned voices as a limited release enabled per workspace — contact Speechify to have it enabled for yours; once enabled every clone you own works there, with no per-voice step. On simba-3.2 a clone is English only, matching the model’s stock voices.
Sample requirements
The voice sample the clone is built from. (The consent recording has its own rules - see Consent.)
Use a good microphone and a quiet room. The clone quality depends directly on sample quality.
Console UI
You can also clone voices through the Speechify Console without writing code:
Import an existing audio file containing the voice
Record directly from your browser
The console walks the same consent step at the end: the speaker reads the phrase into the microphone before the voice is created. A working microphone is required either way, including when the sample itself was uploaded.
The clone option appears once your workspace is on a plan that includes voice cloning and has at least one API key. Using a clone on simba-3.2 needs the per-workspace enablement described above; on simba-3.0, simba-english and simba-multilingual it works as soon as the clone exists.
Sample recording tips
If you need text to read while recording, try this:
Listening is like riding a storytelling rollercoaster, where you can lean back and enjoy the ride without having to steer. The speaker’s voice becomes your trusty guide, leading you through twists and turns. It’s like having a personal audiobook adventure just for you! So, buckle up, and let the fun begin!
For best results:
- Speak naturally at a consistent pace
- Avoid whispering or shouting
- Minimize pauses longer than 2 seconds
- Record in a quiet environment without echo