Speaker Diarization is a pre-trained pipeline available on Hugging Face for determining who spoke when in an audio recording. It forms part of the pyannote.audio toolkit and addresses the task of segmenting and labeling speakers within voice recordings.
The model supports automatic speech recognition workflows through integrated capabilities for speaker diarization, voice activity detection, and overlapped speech detection. It draws on research documented in two arXiv papers from 2021 and 2020. Users invoke it via the pyannote.audio library by loading the pipeline from the repository identifier and passing either a full audio file or a cropped excerpt defined by start and end times.
Delivery occurs as a downloadable model hosted on the Hugging Face platform. The provided code examples demonstrate inference on a local WAV file or on a waveform segment extracted using the accompanying Audio utility. It is distributed under the MIT license, allowing broad reuse in research and development.
The pipeline is intended for developers and researchers working in speech processing who integrate it into local applications, notebooks, or custom audio analysis scripts.
Speaker Diarization is a Voice, TTS & speech product. Automatically identifying and segmenting different speakers in multi-speaker audio recordings. Speaker Diarization is an open-source project aimed at developers. The project is open source (MIT). The product ships for the web, the command line, and API.
pyannote builds and maintains Speaker Diarization, and the product first shipped in 2016. The project is developed in the open on GitHub with 10.3k stars and 13 commits in the last 90 days. Among its 3 catalogued features are speaker diarization, voice activity detection, and overlapped speech detection.
Latest indexed changes and source events
pyannote/speaker-diarization-3.0 verified by the PulseGate indexer
Other apps tracked under the same category.
