OpenAI

OpenAI provides speech-to-text for Vapi voice agents.

You can use OpenAI through Vapi’s default integration, or connect your own account in Integrations.

Supported capability

Speech-to-text

Set transcriber.provider to openai and transcriber.model to a transcription model.

curl -X PATCH "https://api.vapi.ai/assistant/ASSISTANT_ID" \
  -H "Authorization: Bearer $VAPI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "transcriber": { "provider": "openai", "model": "gpt-4o-transcribe" }
  }'

For additional API configuration options, review the OpenAITranscriber fields in the Create Assistant API reference.

Keywords and prompt

Two optional fields guide transcription:

• keywords: An array of words and phrases the model should recognize, such as names, brands, and product terms. Supported only by gpt-transcribe. Each keyword is one line and can’t contain < or >. Up to 16,384 keywords.
• prompt: Free-text context for the transcription, such as the topic of the call or the expected vocabulary. Supported by all three models. Up to 8,192 characters for gpt-transcribe, and 1,024 for gpt-4o-transcribe and gpt-4o-mini-transcribe.

curl -X PATCH "https://api.vapi.ai/assistant/ASSISTANT_ID" \
  -H "Authorization: Bearer $VAPI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "transcriber": {
      "provider": "openai",
      "model": "gpt-transcribe",
      "keywords": ["Vapi", "Acme Dental", "Invisalign"],
      "prompt": "A patient calling a dental office to book or change an appointment."
    }
  }'

For keyword options on other transcribers, see Keywords and keyterm prompting.

Supported languages

The following language values are currently accepted by Vapi for this provider.

Related