Target
EchoSet (Interactive Audio)
Listen to the mixture, then compare SEAL's separation with the clean target. Switch between Speaker A and Speaker B to hear the separated outputs from the same mixture.
Loading audio examples…
Playback notes
Three randomly selected mixtures from EchoSet. Playback and spectrograms are derived directly from the separated waveforms.
Mixture, clean targets and extracted outputs share one gain to prevent clipping. Their players share the playback position. Spectrograms show frequencies with a common magnitude scale.