SPEAKER LAB offline diarization + transcription
GPU · 3080 / 10G OFFLINE

Drop an audio file.

Up to 2 hours. Two-hour files are split into ~5-minute chunks automatically so the 10 GB card (shared with Frigate) never runs out of memory.

1 Diarize — who spoke when (Nemotron 3)
2 Transcribe — chunked ASR (Parakeet 0.6B)
3 Align — words mapped to speakers
4 Review — name speakers, fix conflicts

Working on …

Stagequeued
Elapsed0:00
Audio–
Chunks–

      

Review

…

Export