Are the new Nova and Whisper models available for the streaming (websocket) API?

[davehorton]

Docs are just not clear on this...

[SandraRodgers]

Hi @davehorton ,

Thanks for bringing that up. We'll update the docs with that information.

The Nova model is available with streaming. The code snippets in this Quickstart include model=nova, so you can follow this guide to try it out with the Python or Node SDK. If you're using another SDK, just add model=nova to use our newest most powerful model.

Deepgram's Whisper Cloud is not available with streaming.

Let us know if you need any help getting started.

Sandra

Edit: Here is a link to all the features and which model supports them: https://developers.deepgram.com/documentation/features/

[davehorton]

Hmm, here is the path I am using to test out nova and the deepgram server rejects the connection

/v1/listen?tier=general&model=nova&language=en-US&encoding=linear16&sample_rate=8000

What is wrong? (I am not using python, I am using my own websocket client built in C)

[rastern]

Hi Dave. Try removing "tier=general" from your parameters.

[davehorton]

hmm, even with conversationalai it is taking 3 secs to connect to deepgram and no transcriptions coming back. Is there a system issue today?

[SandraRodgers]

Are you still having problems connecting?

[iamnafets]

I'm having the same issue. Sometimes I get responses, sometimes I don't. Non-deterministic.

[davehorton]

I will have to check later as I am away from my office at the moment. Will update here once I try again.

[SandraRodgers]

Thank you. And we have updated our docs to make it more clear that streaming is not available on Whisper (but is on Nova). We appreciate you calling our attention to that!

[davehorton]

Yes, this is working now.