speaker-diarization

To identify who speaks when in multi-speaker audio, this Skill uses pyannote-audio to assign speech segments to specific speakers. It processes podcasts, interviews, and panels, handling overlapping speech to produce accurate per-speaker diarization data.

Speaker diarization data is exported in JSON, RTTM, and SRT formats. These outputs include per-speaker segmentation and duration data, making the results integration-ready for downstream tasks like transcript alignment and video editing workflows.

Yes, you need a HuggingFace token to run speaker diarization using pyannote-audio. The environment also requires the torch, torchaudio, and speechbrain dependencies to process multi-speaker media and generate per-segment diarization outputs.

Yes, speaker diarization can be used for extracting speaker-specific clips from video. By generating per-speaker segmentation data, the output integrates directly into video editing workflows to isolate and extract individual speaker segments.

Yes, pyannote handles overlapping speech in podcast and interview diarization. It assigns overlapping speech segments to distinct speakers, ensuring accurate speaker identification and per-segment duration tracking in complex multi-speaker media environments.

Using pyannote for speaker identification requires pre-configured torch and speechbrain environments along with a valid HuggingFace token. It is designed for multi-speaker media diarization and does not perform speech-to-text transcription or voice biometric identification.