tuko-tanzwoo/bengali-diarization-natok1

Bengali Speaker Diarization (Natok)

Fine-tuned from pyannote/segmentation-3.0 on Bengali natok audio.

Usage

from pyannote.audio import Model
from huggingface_hub import hf_hub_download

checkpoint = hf_hub_download(
    repo_id="your-username/bengali-diarization-natok",
    filename="pytorch_model.bin",
    token="YOUR_TOKEN"
)
model = Model.load_from_checkpoint(checkpoint)

Training

• Base model: pyannote/segmentation-3.0
• Language: Bengali (Bangla)
• Domain: Natok (drama) audio
• Training files: 10 x ~1hr recordings