[Feature]: deepgram STT as an alternative to openai's whisper

[thiswillbeyourgithub]

The Feature

Using deepgram's api as an endpoint for transcription.

Motivation, pitch

I use the transcription feature or litellm a lot. I think it would be awesome to be able to switch easily from openai to deepgram's stt.

Twitter / LinkedIn details

No response

[krrishdholakia]

can you link me to the endpoint you want us to call? @thiswillbeyourgithub

[thiswillbeyourgithub]

Sure. Here are the most relevant links:

• https://developers.deepgram.com/reference/deepgram-api-overview
  • https://developers.deepgram.com/reference/listen-file
  • https://developers.deepgram.com/reference/listen-live
  • https://playground.deepgram.com/?endpoint=listen&smart_format=true&language=en&model=nova-2
  • https://developers.deepgram.com/docs/migrating-from-openai-whisper-to-deepgram Deepgram really sounds much better than OpenAI regarding audio treatment, at least until we get access to GPT-4o

[AlexM4H]

It could be interesting for realtime STT applications.