VAPI•5d ago

speech context

We’re seeing name misrecognition issues with our Voice AI agent due to phonetically similar names (e.g. “Woosurk Han” being transcribed as “Ushad Khan”).

Does VAPI currently support passing speech contexts / phrase hints / custom vocab to the transcriber?
If yes, what is the correct config field or provider-specific override?

Please advise how we can fix this issues.

Vapi Support BotAPP•12/16/25, 6:36 PM

To address name misrecognition, you can use Deepgram's keyword and keyterm boosting in your Vapi assistant's transcriber config. For example, to boost "Woosurk Han", add:

"transcriber": {
  "provider": "deepgram",
  "model": "nova-2",
  "language": "en",
  "keywords": ["Woosurk", "Han"],
  "keyterm": ["Woosurk Han"]
}

"transcriber": {
  "provider": "deepgram",
  "model": "nova-2",
  "language": "en",
  "keywords": ["Woosurk", "Han"],
  "keyterm": ["Woosurk Han"]
}

This will help the model better recognize the intended name. See the Vapi custom keywords documentation for more details.

Source:

Vapi Custom Keywords Documentation

⤵️ Mention @Vapi Support Bot for follow-ups 🐥

speech context

Similar Threads

Similar Threads

Similar Threads