An unique model for removing crowd sounds from music recordings (applause, clapping, whistling, noise, laugh etc.). Current metrics on our internal dataset for quality control:

Examples of how the model works can be found: here и here.