whisperforge-diarize

Speaker diarization for speech transcription via embedding clustering.

Quick Links

• Full Documentation: WhisperForge Repository
• Architecture: Overview

Features

• Speaker embedding extraction
• Cosine similarity clustering
• SPEAKER_NN label assignment
• Configurable similarity threshold
• Works with SRT and JSON output

Usage

use whisperforge_diarize::DiarizationConfig;
use whisperforge_core::{Model, WhisperConfig};

let config = WhisperConfig::tiny_en();
let model = Model::load(Path::new("models/tiny_en_converted"))?;

// With CLI: use --diarize flag
// wf -a audio.wav -m tiny_en_converted --diarize

CLI Integration

The CLI automatically applies diarization labels when using the --diarize flag:

wf -a audio.wav -m tiny_en_converted --diarize --output-format srt -o output.srt

Output includes speaker labels:

1
00:00:00,000 --> 00:00:05,000
SPEAKER_0: Hello, how are you?

2
00:00:05,000 --> 00:00:10,000
SPEAKER_1: I'm doing great, thanks for asking.

See Also

• whisperforge-core — Library
• whisperforge — CLI binary (wf); wf convert ports HuggingFace safetensors
• whisperforge-align — VAD & SRT

For full documentation, visit the WhisperForge repository.