-
-
Notifications
You must be signed in to change notification settings - Fork 4.6k
Open
Labels
bugSomething isn't workingSomething isn't working
Description
What happened?
When using Deepgram for audio to text transcription, you can specify diarize=true in parameters to get diarization.
In the Deepgram response, the diarized transcript and the classic one are stored in different properties.
The classic one is returned by ['results']['channel'][0]["alternatives"][0]['transcript'], the diarized one is returned by ['results']['channel'][0]["alternatives"][0]['paragraphs']['transcript'].
When I specify the diarize=true parameter, I expect that the text response property contain the diarized transcript and not the classic one as actually (see here).
Relevant log output
curl \
-X POST \
-H "Authorization: Token xxxxxxxxx" \
-H "Content-Type: application/json" \
-d '{"url":"xxxxxxxxxx"}' \
"https://api.deepgram.com/v1/listen?smart_format=true&diarize=true&model=nova-3"
{
"metadata": {
...
},
"results": {
"channels": [
{
"alternatives": [
{
"transcript": "Yep. I said it before, and I'll say it again. Life moves pretty fast. You don't stop and look around once in a while, you could miss it.",
"confidence": 0.99926823,
"words": [
...
],
"paragraphs": {
"transcript": "\nSpeaker 0: Yep. I said it before, and I'll say it again. \nSpeaker 1: Life moves pretty fast. You don't stop and look around once in a while, you could miss it.",
"paragraphs": [
...
],
}
]
}
}
]
}
]
}
}Are you a ML Ops Team?
No
What LiteLLM version are you on ?
v1.79.0
Twitter / LinkedIn details
No response
thomasvielevaneos and RomainJeff
Metadata
Metadata
Assignees
Labels
bugSomething isn't workingSomething isn't working