Self-Supervised Speech Pre-training and Representation Learning Toolkit

[ICLR 2025] SOTA discrete acoustic codec models with 40/75 tokens per second for audio language modeling

[ACL 2024] Official PyTorch code for extracting features and training downstream models with emotion2vec: Self-Supervised Pre-Training for Speech Emotion Representation

A Survey of Spoken Dialogue Models (60 pages)

Ultra-low bitrate speech codec (0.27-1 kbps) with cross-modal alignment and real-time capabilities

A single-layer, streaming codec model providing SOTA audio quality and discrete tokens designed for superior downstream modelability.

MOSS-Audio-Tokenizer is a Causal Transformer-based audio tokenizer built on the CAT architecture. Trained on 3M hours of diverse audio, it supports streaming and variable bitrates, delivering SOTA reconstruction and strong performance in generation and understanding—serving as a unified interface for next-generation native audio language models.

Trainging, inference, and testing of the SAC speech codec model.

This is the code for paper: XY-Tokenizer: Mitigating the Semantic-Acoustic Conflict in Low-Bitrate Speech Codecs

LightHuBERT: Lightweight and Configurable Speech Representation Learning with Once-for-All Hidden-Unit BERT

Survey of audio language models

Official Implementation of Mockingjay in Pytorch

A mini, simple, and fast end-to-end automatic speech recognition toolkit.

Official code for paper:"Speaking Clearly: A Simplified Whisper-Based Codec for Low-Bitrate Speech Coding"

Causal Speech Enhancement Based on a Two-Branch Nested U-Net Architecture Using Self-Supervised Speech Embeddings

DUSTED: Spoken-Term Discovery using Discrete Speech Units

Semi-supervised spoken language understanding (SLU) via self-supervised speech and language model pretraining