Alvor/

audio-tokenizer-demo

Audio samples processed by various neural audio tokenizers for comparison.

	Tokenizers




	

Column Tokens/sec Codebooks

cosyvoice2 25 1

glm4voice 12.5 1

mimoaudio 6.25 8

neucodec 50 1

wavtokenizer 40 1

xcodec2 50 1

unicodec 75 1

bicodec 50 1 (Semantic) + 1 (Global Fixed 35 tokens)

flexicodec <100 1(FSQ)+7(RVQ)

varstok <40 1

MOSS 400 32

H-Codec-1.5 <200 4(Acoustic) + 4(Semantic)