Hello, we recently had a call which wasn't properly transcribed. I mean, it's not even close to how the call went. Please look into it and let us know why that issue happened. Call ID: d4501eb4-c186-47d6-a17e-70a64dc9ba95 Organization ID: 4dd2575c-dec3-4d2f-bc73-63e885ad533c https://storage.vapi.ai/d4501eb4-c186-47d6-a17e-70a64dc9ba95-1751389351539-0f38e9ea-d233-4345-81e0-afa95b90b53d-mono.wav https://cdn.discordapp.com/attachments/1389736155540357243/1389736155989409953/call_transcript.txt?ex=688d4127&is=688befa7&hm=c630f8d0a6099fdf4600aa5ed49b9d531a1c0bf3b79414fc0f2678ef22fe6d87&

Here is a list of recommended changes for your assistant to prevent hallucinations: 1. Reduce Keyterm List (High Priority)

"keyterm": [
  "Ballyhoos", "Wings", "Burger", "Pizza", "Fries", "Beezer"
]

Keep only the 10-15 most essential/unique terms that customers actually struggle to pronounce. 2. Increase Confidence Threshold

"confidenceThreshold": 0.6

This will filter out more uncertain transcriptions. 3. Consider using nova-2 for deepgram transcription. Make these adjustments, run some test calls, and let us know if the results improve

Thank you

Please reach out to us if you have any other questions regarding this issue