Speech-to-text API
Speech-to-text transcription, one API key
Transcribe audio to text fast and accurately, with timestamps and language detection. OpenAI Whisper-compatible, so existing Whisper code works by changing the base URL. Billed per minute of audio, with a multi-route fallback chain so transcription stays up.
POST /v1/audio/transcriptions · $0.50 free credit, no card
2 models, one API key
Switch models with a string. Billed per use, refunded on failure.
Quick start
OpenAI-compatible. Point your client at https://kymaapi.com/v1 with a ky- key.
curl https://kymaapi.com/v1/audio/transcriptions \
-H "Authorization: Bearer YOUR_API_KEY" \
-F model="whisper-v3-turbo" \
-F file="@meeting.mp3"
# Returns: { "text": "..." } (use response_format=verbose_json for timestamps)FAQ
Which transcription models does Kyma support?
Kyma offers Whisper Large v3 Turbo as the default speech-to-text model — fast and inexpensive — plus a premium transcription option for higher accuracy on conversational or code-switching audio. Both use the OpenAI Whisper request shape.
Is the transcription API OpenAI Whisper-compatible?
Yes. Call POST /v1/audio/transcriptions as a multipart upload with a model and file, exactly like the OpenAI audio transcriptions API. Existing Whisper code works by pointing the base URL at https://kymaapi.com/v1.
How much does transcription cost?
Transcription is billed per minute of audio processed, with a one-minute minimum. The default Whisper model costs a fraction of a cent per minute. See the pricing page for exact rates.
Can I get word- or segment-level timestamps?
Yes — request response_format=verbose_json to receive segment timestamps and detected language alongside the transcript.
$0.50 free credit on signup. No credit card required.
Get your free API key →