I am having an issue where my Vapi assistant keeps saying 0 instead of o. For example if I say the last name Moore, it interprets it as M o 0 r e. I've tried everything I know in Vapi API. Deepgram keywords, ElevenLabs pronuciation dictionaries, prompting directly in my assistant prompt. This only happens with double Os. (o o) Call IDs: 1bd09f8c-8b8b-49de-b90a-8ba2adf9d7e9 da3fcd6a-f79f-4d58-bcbb-2d2cd217e3c0 I don't have my own Deepgram API key, I am just using the one available with Vapi and the settings available via Vapi's API. I do not have smart formatting on, nor numerals. No format plan on as well. Please assist as soon as possible. Thank you! — "transcriber": { "model": "nova-3", "keyterm": [ "Bullard", "Tyler", "Longview", "Flint", "..." ], "language": "en", "numerals": false, "provider": "deepgram", "endpointing": 200, "confidenceThreshold": 0.70, "mipOptOut": true }

Thanks for sharing the details and call IDs. In this case, we recommend two adjustments to improve how “oo” sounds are handled: 1. Enable

smartFormat

– this helps the model distinguish between letters and digits more reliably. 2. Update your

formatPlan

– specifically remove

"number"

from

formatPlan.formattersEnabled

, so the assistant won’t try to interpret letter sequences as digits. After making these changes, test again with the same examples (like “Moore”) and let us know if the issue persists.

Okay, thanks I'll make the adjustments and test