TIL: Clone a voice in one API call with Speechify

TIL Tuesday: the Speechify voice cloning API does it all in a single POST. You send a 10-second audio clip plus a consent record, and it hands back a voice ID you can use straight away. No dashboard, no waiting, no separate fine-tuning step.

I expected voice cloning to involve uploading, waiting for a model to train, and checking a dashboard later. I was wrong. The voice cloning endpoint is a multipart POST to /v1/voices. You include the audio file plus a consent field containing a JSON object with fullName and email of the person whose voice you’re cloning. The response is a voice_id you can use immediately on any speech endpoint, same as a catalog voice.

What surprised me most: there is no separate model fine-tuning step. The clone is instant. Zero-shot means the API synthesizes in that voice from the first request, no training wait, no “your voice is being prepared” page.

The audio needs to be between 10 and 30 seconds. Clean audio with one speaker and no background noise gives the best result, but I tested it with a phone recording and it still captured the speaker’s cadence well enough to be recognizable. Not studio quality, but usable.

Once you have the voice_id, pass it as voice_id on POST /v1/audio/speech or POST /v1/audio/stream with model: "simba-english" or model: "simba-multilingual". The cloned voice speaks whatever text you give it, in the language of the model you pick.

FAQ

How long does Speechify voice cloning take?

The API returns a voice ID in the same response as the upload. There is no separate processing step. The clone is available for synthesis immediately after the POST returns.

What audio format does Speechify voice cloning accept?

The endpoint accepts common audio formats as a binary upload. The sample should be 10 to 30 seconds long, under 5MB, with one speaker and minimal background noise.

Can I use a cloned voice across multiple languages?

Yes. Cloned voices on simba-multilingual can synthesize text in any language the model supports. The clone captures the speaker’s tone and cadence but the accent applies to the target language.

Do I need permission to clone someone’s voice?

Yes. The API requires a consent field containing the speaker’s full name and email as a JSON object. Speechify stores these with the voice record for provenance.