Nova-3 sometimes skips full sentences in transcription output

[M-tarek93]

Hi Deepgram team 👋

I ran into a reproducible issue where an entire sentence appears to be missing from the transcription output. the sentence is clear and audible. also can be transcribed without issues when using Nova-2 or even Nova-3 but with multilingual

Details

Here’s the request I’m making:

const result = await client.listen.prerecorded.transcribeUrl({
  url,
  model: 'nova-3',
  language: 'en',
  smart_format: true,
  utterances: true
});

In the returned transcript, a full sentence around second 23 is completely dropped. The audio clearly contains speech in that segment, but it does not appear in the output at all (not even partially or as low-confidence text).

Request ID: 019db972-1d5e-7091-ab0d-f81418fecf6f

Expected Behavior

The sentence to be transcribed (even if imperfect)

…but not a complete omission.

Questions

Is this a known limitation or behavior of nova-3?
Are there recommended parameters or settings to reduce the likelihood of dropped segments?

Additional Context

I’ve noticed that it's being transcribed correctly with the same audio file in these 2 cases

• using Nova-2 (Request ID: 019db97c-5b31-7c90-84c3-fe827f88774c)
• using Nova-3 but with language: multi (Request ID: 019db97d-548b-7fd0-9eeb-4a292e56dc9e)

Happy to provide the audio sample if that helps debug.

Thanks in advance!

[deepgram-community[bot]]

Thanks for asking your question. Please be sure to reply with as much detail as possible so the community can assist you efficiently.
_{Consider joining our Discord community for more opportunity to engage with your fellow Deepgram users. You can earn points which can be redeemed for cool stuff by being active in our communities!}

[deepgram-community[bot]]

Hey there! It looks like you haven't connected your GitHub account to your Deepgram account. You can do this at https://community.deepgram.com - being verified through this process will allow our team to help you in a much more streamlined fashion.

[deepgram-community[bot]]

It looks like we're missing some important information to help debug your issue. Would you mind providing us with the following details in a reply?

• The deepgram product you are using (e.g Speech to Text, Agent API)

[M-tarek93]

Speech to Text

[deepgram-community[bot]]

Hiya~

I'd be happy to help you look in to this! A couple of things:

1. Yes, the audio sample would be very helpful! If it is not something you can share in a public thread, you can email it to me at [naomi.carrigan@deepgram.com](mailto:naomi.carrigan@deepgram.com)
2. Do you happen to have a minimal reproduction example we can run?

This message was sent by Naomi from Deepgram, via our community automation.

[M-tarek93]

Sorry, I wasn’t notified about the last comment.

Sure, the audio is part of a public YouTube video.
Video link: https://roleplay-dev-1.s3.eu-central-1.amazonaws.com/video_54mb.mp4

You can use the provided snippet with this link.

Thanks in advance!