tool

pyannote speaker identification tool.

Functions

cropped_waveform

Crops an audio track and returns its corresponding waveform.

Classes

PyannoteIdentificationTool

pyannote speaker identification tool.

class PyannoteIdentificationTool(model_names, api_token=None, device='cpu', overwrite=False, verbose=True)

Bases: IdentificationTool

pyannote speaker identification tool.

Parameters:

• model_names (Sequence[str]) – The names of the models to use.

• api_token (Optional[str]) – The HuggingFace API token to use.

• device (str) – The device where the computation should be executed.

• overwrite (bool) – Whether to overwrite existing files, otherwise raise an error.

• verbose (Union[bool, int]) – Whether to execute the computation verbosely.

inference(mixed_audio_path, diarization_path, mono_audio_paths, identification_path)

pyannote’s backed inference method.

Parameters:

• mixed_audio_path (Union[str, Path]) – Path to the mixed audio track.

• diarization_path (Union[str, Path]) – Path to the diarization file.

• mono_audio_paths (Sequence[Union[str, Path]]) – Path to the mono audio tracks.

• identification_path (Union[str, Path]) – Path to the identification file.

Returns:

cropped_waveform(path, start, end, sample_rate=32000)

Crops an audio track and returns its corresponding waveform.

Parameters:

• path (Union[str, Path]) – Path to the audio track.

• start (float) – Start of segment in seconds.

• end (float) – End of the segment in seconds.

• sample_rate (int) – Sample rate of the audio track.

Return type:

Tensor

Returns:

Tensor containing the waveform of the audio segment.