I have a use case where AI has to collect the email id & name from the user. When the user spells s t e p h a n i e, AI takes it as s t e b h a n i e. If the pronounciation very slightly diffent it takes the wrong spelling. Email id collection is sensitive. The same goes for name as well. How to fix this? I think this cannot be fixed with prompt as it is done by the transcriber

Thank you for sharing the details. To help you further, could you please let me know which transcriber provider (e.g., AssemblyAI, Deepgram, Gladia) you are currently using in your Vapi AI setup? Sources: - [Best Practices for Keyword and Keyterm Boosting](https://docs.vapi.ai/customization/custom-keywords) - [AssemblyAI Transcriber Provider Info](https://docs.vapi.ai/providers/transcriber/assembly-ai) - [Custom Transcriber Integration](https://docs.vapi.ai/customization/custom-transcriber) - [Changelog: Enhanced Transcription Features](https://docs.vapi.ai/changelog/2025/9/8)

Im using Deepgram nova 2. Keyword boosting will not help. I cannot put all emails and name in keyword boosting

Hey! To help track down this issue, could you share: \- The call ID \- When exactly this happened (the timestamp) \- What response you expected to get \- What response you actually got instead This would really help us figure out what went wrong!

019a24bf-062e-7ee5-b7d8-060ec32c00d3 - Please check this call id. There are multiple mistakes in the transcript. When the user says Vice President it takes it as voice present. Exact timing is 1:48:59 PM(+00:43.52). When User says his name as Ajith it takes it as Agent.

Hey, thanks for sharing your details. Our team will look into it and get back to you soon with an update.

Praveen, this is urgent. We are getting a lot of questions on the quality.

Please increase your maxTokens to 500 or switch to a different LLM like gpt-4.1 mini or gpt-4o.

This issue exists with all LLMs. It is the problem with STT (Speech to text) transcribers

Switch the language on the transcriber to

en-IN

or use AssemblyAI for

en-IN

Same issue. for

DE

language mainly. Transcribers cannot get emails correctly

Please provide a call id and we will take a look at it