Kimi-Audio Technical Report

KimiTeam, Ding Ding, zeqian ju, Yichong Leng, liusongxiang, Tong Liu, shangzeyu, Kai Shen, Wei Song, Xu Tan, Heyi Tang, zhengtao, Chu Wei, Yifei Xin, Xu Xinran, Jianwei Yu, Yutao Zhang, Xinyu Zhou, Y. Charles, Jun Chen, Yanru Chen, Du, Weiran He, Zhenxing Hu, Guokun Lai, Qingcheng Li, Yangyang Liu, Weidong Sun, Jianzhou Wang, yuzhi wang, Yuefeng Wu, Yuxin Wu, Dongchao Yang, Hao Yang, Ying Yang, ZHILIN YANG, aoxiong, Ruibin Yuan, Yutong Zhang, zhouzaida

Kimi-Audio Technical Report: 21 upvotes on Hugging Face Daily Papers, #7 of 12 papers on 2025-04-28. Day-by-day upvote history.

We present Kimi-Audio, an open-source audio foundation model that excels in audio understanding, generation, and conversation. We detail the practices in building Kimi-Audio, including model architecture, data curation, training recipe, inference deployment, and evaluation. Specifically, we leverage a 12.5Hz audio tokenizer, design a novel LLM-based architecture with continuous features as input and discrete tokens as output, and develop a chunk-wise streaming detokenizer based on flow matching. We curate a pre-training dataset that consists of more than 13 million hours of audio data covering a wide range of modalities including speech, sound, and music, and build a pipeline to construct high-quality and diverse post-training data. Initialized from a pre-trained LLM, Kimi-Audio is continual pre-trained on both audio and text data with several carefully designed tasks, and then fine-tuned to support a diverse of audio-related tasks. Extensive evaluation shows that Kimi-Audio achieves state-of-the-art performance on a range of audio benchmarks including speech recognition, audio understanding, audio question answering, and speech conversation. We release the codes, model checkpoints, as well as the evaluation toolkits in https://github.com/MoonshotAI/Kimi-Audio.

Paper page on Hugging Face · arXiv

Data: hysts-bot-data/daily-papers-stats and the Daily Papers API. Open data: tardellirs/paper-pulse-data. Sister project: Model Pulse, the download history of every model on the Hub.