Separation

Singing voice extraction

Leakage-aware Carnatic Singing Voice Separation

Note

REQUIRES: tensorflow

class compiam.separation.singing_voice_extraction.cold_diff_sep.ColdDiffSep(model_path=None, download_link=None, download_checksum=None, gpu='-1')[source]

Leakage-aware singing voice separation model for Carnatic Music.

download_model(model_path=None, force_overwrite=False)[source]

Download pre-trained model.

separate(input_data, input_sr=44100, clusters=5, scheduler=4, chunk_size=3, gpu='-1')[source]

Separate singing voice from mixture.

Parameters:

  • input_data – Audio signal to separate.

  • input_sr – sampling rate of the input array of data (if any). This variable is only relevant if the input is an array of data instead of a filepath.

  • clusters – Number of clusters to use to build the separation masks.

  • scheduler – Scheduler factor to weight the clusters to be more or less restirctive with the interferences.

  • gpu – Id of the available GPU to use (-1 by default, to run on CPU), use string: ‘0’, ‘1’, etc.

Returns:

Singing voice signal.