OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Audio & Voice AI

Speaker Change Detection

Speaker Change Detection is an open-source research implementation for detecting speaker-change timestamps and segmenting conversation audio, built on a speaker-segmentation method and modified for speaker-change detection.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Speaker Change Detection — OpenViST project profile in Audio & Voice AI.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when you want to inspect or adapt a research implementation that covers both speaker-change detection and speaker segmentation, rather than a system that identifies each speaker throughout a recording.

Who it is for

  • Researchers studying speaker-change detection or speaker segmentation.
  • Developers preparing speech datasets or exploring speaker-turn preprocessing.
  • Practitioners who need file-level inference for speaker-change or segmentation experiments.

Documented uses

  • Identify timestamps where the active-speaker configuration changes.
  • Segment conversation audio into speaking turns, including overlapping speech probabilities.
  • Prepare AMI Meeting Corpus data, train a model, and run inference on an audio file.
Before you choose

Requirements & limits

The code is publicly available under the GPL-3.0 license.

  • Speaker-change detection marks changes in the active-speaker configuration; segmentation identifies speaking activity and turns.
  • Segmentation can represent multiple simultaneous speakers but does not provide speaker identity across the full file.
  • The README reports that the tested Transformer performed worse than the LSTM in its experiments.
  • Training uses PyTorch Lightning and the README describes training on a GPU.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Run inference on an audio file

Use the repository's documented file-processing script with a model checkpoint and audio input.

Before you start

  • A model checkpoint
  • An audio file
  • The repository environment installed with Poetry

Steps

  1. Open process_file.py and replace short_scd_bigdata.ckpt with the path to the checkpoint.

  2. Replace test_audio_similar.wav with the path to the input audio file.

  3. Set DO_SCD to match the task configuration used for training.

  4. Run the file-processing script using the installed project environment.

Expected outcome

The script processes the audio file with the selected inference configuration.

Guide limitations

  • The README does not specify a separate output format in the documented steps.
  • Speaker segmentation does not assign speaker identities across the full recording.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
15 October 2026
Original website title
HHousen/speaker-change-detection · Official homepage title · retained in its source language
How we review sourcesSuggest a correction