Speaker Diarization using OpenAI Whisper and Pyannote

Table of content

• Introduction
• Prerequisties
• Docker Setup
• Usage
• Screenshots

Introduction

Speaker Diarization pipeline based on OpenAI Whispe and Pyannote.

Prerequisties

• Docker==20.10.7
• Nvidia-Docker

Docker Setup

I prefer using the Docker because of simplicity. You just need to run the gpu-enabled docker container and everything is setup for you

git clone https://github.com/leviethung2103/whisper_speaker_diarization
cd whisper_speaker_diarization
docker run --gpus all -d -it -p 8848:8888 -v $(pwd):/home/jovyan/work -e GRANT_SUDO=yes -e JUPYTER_ENABLE_LAB=yes --user root cschranz/gpu-jupyter:v1.4_cuda-11.6_ubuntu-20.04 

Usage

• Access the jupyter lab via http://localhost:8848
• Start with jupyter notebook 01_Speaker_Diarizateion.ipynb.
• Default password is gpu-jupyter

Screenshots

Here is the output after running modules.