mispeech/dashengtokenizer

Audio To Audio·mispeech· 2.7K· 12

transformers apache-2.0 Audio To Audio 805.0M params arxiv:2602.23765arxiv:2602.2602license:apache-2.0region:us

DashengTokenizer is a high-performance continious audio tokenizer designed for audio understanding and generation tasks. Compared to previous works, our framework trains a single linear layer to enable audio generation for semantically strong encoders.

Open in MLForge Sign up free Desktop app Source ↗

# pull & run locally
pip install mlforge-sdk && mlforge pull mispeech/dashengtokenizer

Model details

Task

Audio To Audio

Provider

mispeech

Framework

transformers

Parameters

805.0M

Size

3.0 GB

License

apache-2.0

Downloads

2.7K

Likes

Paper

arXiv:2602.23765

Updated

2026-04-21

About mispeech/dashengtokenizer

Related Audio To Audio

B nvidia/bigvgan_v2_22khz_80band_256x Audio To Audio 1.6M 29 🤗 HF B nvidia/bigvgan_v2_44khz_128band_512x Audio To Audio 500.6K 75 🤗 HF D aufklarer/DeepFilterNet3-CoreML Audio To Audio 210.5K 2 🤗 HF P aufklarer/PersonaPlex-7B-MLX-4bit Audio To Audio 206.4K 37 🤗 HF N neuphonic/neucodec Audio To Audio 188.6K 108 🤗 HF T JusperLee/TIGER-DnR Audio To Audio ·4.2M params 145.4K 23 🤗 HF

Browse all Audio To Audio models →