Hey guys, I use nova3-medical and see speaker 0 always in live mode. Python 3.12. new Deepgramm SDK
my options is
model="nova-3-medical",
diarize=True,
smart_format=True,
language="en",
interim_results=True,
utterance_end_ms="1000",
sample_rate=16000,
encoding="linear16",
channels=1,
vad_events=True
)```
logs is ``` {
"channel": {
"alternatives": [
{
"transcript": "Good morning. I'm doctor Shen. Must be Karen. Yes. Before we",
"confidence": 0.9802246,
"words": [
{
"word": "good",
"start": 1.5999999,
"end": 1.92,
"confidence": 0.9819336,
"punctuated_word": "Good",
"speaker": 0
....
{
"word": "yes",
"start": 4.66,
"end": 5.06,
"confidence": 1.0,
"punctuated_word": "Yes.",
"speaker": 0
},
..."metadata": {
"model_info": {
"name": "medical-nova-3",
"version": "2025-03-04.1",
"arch": "nova-3"
},
"request_id": "8cd62e4c-678b-41d4-a181-276ababe5f30",
"model_uuid": "7365352c-4d5c-41e8-8f72-27770c4cf3a3"
},```
So after word, Karen had another voice. But Deepgram sees it as the same. Transcription is true. But diarization is not.