AI voiceover with my own voice: clone once, reuse in TTS

To make an AI voiceover in your own voice, clone your voice once from a short recording, then choose that clone as the voice each time you turn a script into speech. You record one sample instead of every line. In the Sume app today, Assets → Voices clones a voice from a clip you upload or record; the voice's voi_ id then goes in voice.id on POST /v1/tts-1.0/generate, or you tag the voice with @ in a Sume agent chat.

Cloning happens in the Sume app; the public API has no route that creates a voice. The app steps below are read from the Sume app's current code, and the speech facts from the TTS 1.0 schema in the Sume API reference, the OpenAPI document behind the API reference docs, read on 2026-09-27.

How do I clone my voice in Sume?

• Open Assets → Voices, choose Add a new voice, then Clone a voice (“Upload or record audio”).
• Give the voice a name, which is required, and pick its gender and the language it speaks.
• Upload a clip or record one in the browser. The limits are in the table below.
• When the voice is ready, open its More menu and choose Copy ID to copy its voi_ id.
• Only clone your own voice, or one you have permission to use.

How do I use the cloned voice for a voiceover?

Send the id as voice.id. TTS 1.0 resolves a Voices library id to the stored voice before the job is queued. In the current code, an id that is missing, not ready yet, or not available to your workspace fails with 400 invalid_voice_id, and no job is created. Text to speech API covers the other request fields.

In a Sume agent chat today, you can tag a saved voice with @ and ask for the lines; the agent's current instructions tell it to prefer the tagged voice for speech in that turn.

curl -X POST https://api.sume.com/v1/tts-1.0/generate \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: promo-voiceover-v1" \
  -d '{
    "transcript": "Hi, I am the founder. Here is what we built this year.",
    "voice": { "id": "voi_…" },
    "language": "en"
  }'

Can my cloned voice speak another language?

You can ask it to, after a double-check. Each voice currently records one of the 16 languages in the table, and the TTS request's language sets the language of the speech; set it for every non-English script, because an omitted language defaults to English.

If the voice's recorded language differs from the request's, the API currently answers 409 tts_voice_language_mismatch before any job or charge, and its message warns that pronunciation may sound unnatural. After you confirm, resend the same request with confirm_language_mismatch: true; confirming does not change the voice or the language.

What does speech in my cloned voice cost?

Speech in the cloned voice bills like any TTS job: $0.0475 per 1,000 characters, plus a 5.5% agent fee by default, spaces and punctuation included, for up to 20,000 characters per request. The output is MP3 by default, or WAV if you set output_format.

To put the voice on a face, lip-sync a still image to the TTS audio; how to clone yourself with AI covers that path.