Audio samples for "SoundStream: An End-to-End Neural Audio Codec"

Authors: Neil Zeghidour, Alejandro Luebs, Ahmed Omran, Jan Skoglund, Marco Tagliasacchi

About: The audio examples in this page have been randomly selected from the evaluation splits of SoundStream, including clean speech, noisy speech, and music.

Low Bitrate

	Reference	SoundStream (3kbps)	SoundStream (scalable, 3kbps)	Lyra (3kbps)	Opus (6kbps)	EVS (5.9kbps)	EVS (9.6kbps)	Opus (12kbps)
Clean speech
Noisy speech
Music

Medium Bitrate

	Reference	SoundStream (6kbps)	SoundStream (scalable, 6kbps)	EVS (5.9kbps)	Opus (6kbps)	EVS (13.2kbps)	Opus (16kbps)
Clean speech
Noisy speech
Music

High Bitrate

	Reference	SoundStream (12kbps)	SoundStream (scalable, 12kbps)	Opus (12kbps)	EVS (13.2kbps)	EVS (16.4kbps)	Opus (20kbps)
Clean speech
Noisy speech
Music

Denoising (6kbps)

Input	Target	SoundStream (denoising off)	SoundStream (denoising on)