Opensource Audio Datasets

[freelerobot]

Dataset	Audio Type	Source
llama3-s dataset	synthetic speech (Q&A)	Homebrew
LAION-audio-630k	speech, nature
Clotho	speech	Tampere
AS-20K AudioSet	speech	Youtube
ESC-50 Environmental Sound Classification	environment, noise
VoxCeleb		Youtube
VoxLingua		Youtube
LibriSpeech	Audiobooks	OpenSLR
CommonVoice	Speech	Mozilla
Fleurs
SpeechInstruct
Gigaspeech	speech	Youtube
Babel
YODAS manual/automatic	speech	Youtube
MSR-86k	speech
Meld	emotion
MusicCaps	music

[freelerobot]

Also Qwen benchmarks: https://github.com/QwenLM/Qwen2-Audio