You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

The collected information will help acquire a better knowledge of userbase and help its maintainers apply for grants to improve it further. If you are an academic researcher, please cite the relevant papers in your own publications using the model. If you work for a company, please consider contributing back to development (e.g. through unrestricted gifts). We also provide scientific consulting services around speaker diarization and machine listening.

Log in or Sign Up to review the conditions and access this model content.

Using this open-source model in production?
Make the most of it thanks to our consulting services.

🎹 Speaker segmentation

Paper | Demo | Blog post



Relies on 2.1.1: see installation instructions.

# 1. visit and accept user conditions
# 2. visit to create an access token
# 3. instantiate pretrained model
from import Model
model = Model.from_pretrained("pyannote/segmentation", 

Voice activity detection

from import VoiceActivityDetection
pipeline = VoiceActivityDetection(segmentation=model)
  # onset/offset activation thresholds
  "onset": 0.5, "offset": 0.5,
  # remove speech regions shorter than that many seconds.
  "min_duration_on": 0.0,
  # fill non-speech regions shorter than that many seconds.
  "min_duration_off": 0.0
vad = pipeline("audio.wav")
# `vad` is a pyannote.core.Annotation instance containing speech regions

Overlapped speech detection

from import OverlappedSpeechDetection
pipeline = OverlappedSpeechDetection(segmentation=model)
osd = pipeline("audio.wav")
# `osd` is a pyannote.core.Annotation instance containing overlapped speech regions


from import Resegmentation
pipeline = Resegmentation(segmentation=model, 
resegmented_baseline = pipeline({"audio": "audio.wav", "baseline": baseline})
# where `baseline` should be provided as a pyannote.core.Annotation instance

Raw scores

from import Inference
inference = Inference(model)
segmentation = inference("audio.wav")
# `segmentation` is a pyannote.core.SlidingWindowFeature
# instance containing raw segmentation scores like the 
# one pictured above (output)


  Title = {{End-to-end speaker segmentation for overlap-aware resegmentation}},
  Author = {{Bredin}, Herv{\'e} and {Laurent}, Antoine},
  Booktitle = {Proc. Interspeech 2021},
  Address = {Brno, Czech Republic},
  Month = {August},
  Year = {2021},
  Title = {{ neural building blocks for speaker diarization}},
  Author = {{Bredin}, Herv{\'e} and {Yin}, Ruiqing and {Coria}, Juan Manuel and {Gelly}, Gregory and {Korshunov}, Pavel and {Lavechin}, Marvin and {Fustes}, Diego and {Titeux}, Hadrien and {Bouaziz}, Wassim and {Gill}, Marie-Philippe},
  Booktitle = {ICASSP 2020, IEEE International Conference on Acoustics, Speech, and Signal Processing},
  Address = {Barcelona, Spain},
  Month = {May},
  Year = {2020},

Reproducible research

In order to reproduce the results of the paper "End-to-end speaker segmentation for overlap-aware resegmentation ", use pyannote/segmentation@Interspeech2021 with the following hyper-parameters:

Voice activity detection onset offset min_duration_on min_duration_off
AMI Mix-Headset 0.684 0.577 0.181 0.037
DIHARD3 0.767 0.377 0.136 0.067
VoxConverse 0.767 0.713 0.182 0.501
Overlapped speech detection onset offset min_duration_on min_duration_off
AMI Mix-Headset 0.448 0.362 0.116 0.187
DIHARD3 0.430 0.320 0.091 0.144
VoxConverse 0.587 0.426 0.337 0.112
Resegmentation of VBx onset offset min_duration_on min_duration_off
AMI Mix-Headset 0.542 0.527 0.044 0.705
DIHARD3 0.592 0.489 0.163 0.182
VoxConverse 0.537 0.724 0.410 0.563

Expected outputs (and VBx baseline) are also provided in the /reproducible_research sub-directories.

Downloads last month
Inference API has been turned off for this model.

Spaces using pyannote/segmentation 14