loud.cpp

Whisper.cpp with diarization

Features

• Diarize with onnxruntime
• Segment using pyannote-audio model
• Transcribe using OpenAI whisper
• Support macOS, Windows, Linux
• Comes with FFMPEG
• Support any audio or video format
• Download models automatically with --setup
• Colored output for easy identification of speakers

Install

Download and extract it from releases

Usage

./loud test.wav --json transcript.json --setup

Building

See building.md