cicora.ai
English
API specificationSpeech

Audio

Use separate operations for speech synthesis, transcription, and audio input to a model.

Integration reference: choose parameters and available capabilities from your account API settings and the model catalogue for your environment.

Synthesize speech

For narration, send text, model, and a voice available to the selected model. Handle the response as a binary audio stream, not a JSON string.

Assign a file extension and Content-Type from the actual response format before saving or playing it.

Text-to-speech request
bash
curl --request POST https://cicora.ai/api/v1/audio/speech \
  --header "Authorization: Bearer $CICORA_API_KEY" \
  --header 'Content-Type: application/json' \
  --output narration.mp3 \
  --data '{
    "model": "mistralai/voxtral-mini-tts-2603",
    "voice": "en_paul_neutral",
    "input": "Your delivery is scheduled for tomorrow.",
    "response_format": "mp3"
  }'

Transcription and input

  • Send an audio file to /audio/transcriptions in its documented format.
  • Limit file duration and size on your side before upload begins.
  • Keep language, timestamps, and segments only if the selected operation returns them.

Quality and availability

Voice, format, languages, and synchronicity depend on the model. Show a user only controls confirmed by the catalogue.

For longer files, create background work in your product so a network timeout does not define the user workflow.

Continue with