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

🚀🚀 「大模型」2小时完全从0训练26M的小参数GPT！🌏 Train a 26M-parameter GPT from scratch in just 2h!

LongCat Audio Tokenizer and Detokenizer

The best ChatGPT that $100 can buy.

PyTorch media decoding and encoding

Game-Time: Evaluating Temporal Dynamics in Spoken Language Models

Python - 100天从新手到大师

We Speech Toolkit, LLM based Speech Toolkit for Speech Understanding, Generation, and Interaction

🔥 The Web Data API for AI - Turn entire websites into LLM-ready markdown or structured data

Data Pipeline, Models, and Benchmark for Omni-Captioner.

Qwen3-omni is a natively end-to-end, omni-modal LLM developed by the Qwen team at Alibaba Cloud, capable of understanding text, audio, images, and video, as well as generating speech in real time.

MiMo: Unlocking the Reasoning Potential of Language Model – From Pretraining to Posttraining

🥢像老乡鸡🐔那样做饭。主要部分于2024年完工，非老乡鸡官方仓库。文字来自《老乡鸡菜品溯源报告》，并做归纳、编辑与整理。CookLikeHOC.

Official code of ICML 2025 paper "NTPP: Generative Speech Language Modeling for Dual-Channel Spoken Dialogue via Next-Token-Pair Prediction"

A complete computer science study plan to become a software engineer.

HighRateMOS is the first non-intrusive MOS prediction model that explicitly models sampling rates, achieving first place in five out of eight metrics in AudioMOS Challenge 2025 Track3.

Official baseline for ICASSP 2026 URGENT Challenge Track 2 (Speech Quality Assessment)

[ICLR 2025] Official PyTorch Implementation of Gated Delta Networks: Improving Mamba2 with Delta Rule

This repository provides a benchmark for prompt injection attacks and defenses

Model analyzer in PyTorch

(ICASSP 2025, official code)FlowSE: Flow Matching-based Speech Enhancement

Extract phoneme-level timestamps from speeh audio.

A Benchmark for Evaluating Turn-Taking and Overlap Handling in Full-Duplex Spoken Dialogue Models

SCOREQ: Speech COntrastive REgression for Quality Assessment (NeurIPS 2024)

Official implementation of the paper "Pre-trained Speech Processing Models Contain Human-Like Biases that Propagate to Speech Emotion Recognition"

Make text LLMs listen and speak