Voice Clone (MiniMax)

Clone a voice from a 10-second to 5-minute reference recording. Returns a voice_id usable in /v1/audio/speech with any MiniMax HD or Turbo SKU.

POST /v1/audio/voice-clone

Voice Clone (MiniMax)

cURL

curl --request POST \
  --url https://kymaapi.com/v1/audio/voice-clone \
  --header 'Authorization: Bearer <token>' \
  --header 'Content-Type: application/json' \
  --data '
{
  "name": "<string>",
  "model": "<string>"
}
'

Synchronous endpoint. Upload a reference audio clip via multipart form, get back a voice_id you can pass to /v1/audio/speech on any MiniMax voice model.

curl -X POST https://kymaapi.com/v1/audio/voice-clone \
  -H "Authorization: Bearer $KYMA_API_KEY" \
  -F file=@reference.mp3 \
  -F name="brand-narrator"

Request

multipart/form-data body.

file · file · required

Reference audio. MP3, WAV, or M4A. Max 10 MB. Duration must be 10 seconds to 5 minutes — Kyma parses the audio header locally and rejects out-of-range files with 400 reference_too_short or 400 reference_too_long before charging. Longer clips don’t improve clone quality and waste upload bandwidth.

name · string

Optional human-readable label, max 64 chars. Surfaced in the response and stored alongside the ownership row for your reference.

model · string · default: "minimax-voice-clone"

Voice clone SKU. Currently only minimax-voice-clone is supported.

Response

200 OK JSON.

{
  "voice_id": "kyma_3f8e2a1b4c9d7e60",
  "name": "brand-narrator",
  "model": "minimax-voice-clone",
  "cost_usd": 2.10,
  "balance_usd": 47.90
}

Field	What
voice_id	Use this in /v1/audio/speech voice_id field. Namespaced as kyma_<rand>.
name	Echo of the label you sent (or null).
cost_usd	Flat charge applied ($2.10).
balance_usd	Remaining balance after settle.

X-Kyma-Model, X-Kyma-Cost-USD, and X-Kyma-Balance-USD headers are also set.

Pricing

Flat $2.10 per cloned voice. One-time charge — once cloned, the voice_id is reusable in unlimited TTS calls.

Ownership

Cloned voice IDs are gated per Kyma user. If user A passes user B’s voice_id to /v1/audio/speech, the request returns 403 voice_not_owned. Voice IDs that aren’t on file are assumed to be MiniMax system voices (browseable by everyone).

Errors

Status	error.code	When
400	not_a_voice_clone_model	model is not a clone SKU
400	invalid_request	missing file, invalid form data
400	reference_too_short	audio duration < 10 seconds
400	reference_too_long	audio duration > 5 minutes
400	audio_unreadable	could not parse duration — file may be corrupt or in an unsupported codec
402	insufficient_credits	balance below $2.10
413	invalid_request	audio file > 10 MB
415	invalid_request	audio format not MP3/WAV/M4A
500	ownership_write_failed	clone succeeded but ownership row insert failed (no charge applied; safe to retry)
502	provider_error	upstream MiniMax failure

See also

• POST /v1/audio/voice-design — generate a voice from a written description (no reference audio)
• POST /v1/audio/speech — use the cloned voice
• Voice Clone (model) — service overview