ComfyUI_Speaker_Diarization

Custom node for ComfyUI that performs speaker diarization with chronological output ordering.

By ldfla·Created 4 months ago·Updated 4 months ago· 0

ldfla/ComfyUI_Speaker_Diarization

Nodes1

On cloudLocal install

CategoryAudio/Isolation

Stars0

Updated4 months ago

Nodes (1)

Speaker Diarizer (Up to 5 Speakers)

Audio/Isolation

Readme

ComfyUI Pyannote Speaker Diarizer

Custom node for ComfyUI that performs speaker diarization with chronological output ordering.

Features

Up to 5 speaker outputs
Chronological mapping:
- speaker_1_audio = first speaker to talk
- speaker_2_audio = second speaker to talk
- speaker_3_audio = third speaker to talk
- speaker_4_audio = fourth speaker to talk
- speaker_5_audio = fifth speaker to talk
Useful for telephony / call-center / IVR / conference audio
Returns:
- 5 isolated speaker audio tracks
- human-readable summary
- JSON with diarization segments

Requirements

ComfyUI
Python environment with:
- pyannote.audio >= 3.1
- torch
- torchaudio
Hugging Face token with access to:
- pyannote/speaker-diarization-3.1

Installation

Clone inside your ComfyUI custom_nodes folder:

cd ComfyUI/custom_nodes
git clone https://github.com/ldfla/ComfyUI-Speaker-Diarization.git
install requirements:

cd ComfyUI/custom_nodes/ComfyUI-Speaker-Diarization
pip install -r requirements.txt