Voice design API

Send POST /voices with voices:write scope and kind: "design".

curl --fail-with-body https://app.verbatik.com/api/v1/voices \
  -H "Authorization: Bearer $VERBATIK_API_KEY" \
  -H 'Content-Type: application/json' \
  -H 'Idempotency-Key: designed-narrator-001' \
  -d '{
    "kind":"design",
    "name":"Warm narrator",
    "description":"A warm adult narrator with a clear British accent and measured pacing.",
    "preview_text":"Every object has a story. Here is ours."
  }'

name is required, with a maximum of 80 characters. description is required, with a maximum of 1,000 characters. Optional preview_text is limited to 500 characters.

Describe vocal qualities rather than requesting an unauthorized imitation of a real person. Listen to the preview before adopting the voice for a production.

Successful creation returns 201 with a voice resource ID and the charged credits. Creation and preview are billed together; later speech generation is a separate operation. Use the ID with the Speech Cloning model in POST /speech.

Use the same idempotency key and request body when retrying after a timeout. Give a genuinely new voice request a new key.