2025 LRAC Challenge Results
Table of Contents
Introduction
Thank you to all those who participated in the 2025 LRAC Challenge!
The crowdsourced listening tests for the test phase of challenge have now concluded! 🎉
All submissions were evaluated using blind crowdsourced listening tests covering transparency, intelligibility, and robustness to noise/reverb (MUSHRA, MOS/DMOS, DRT) as outlined in the Evaluation section. Systems were required to meet LRAC’s low resource constraints on bitrate, latency, and compute outlined in the Official Rules.
Challenge Participation
- More than 60 participants from 11 countries registered for the challenge and formed 26 teams (24 in Track 1, and 22 in Track 2) across academia, industry, and individual contributors.
- Countries represented: Australia, Bahrain, China, Czechia, France, India, Japan, Poland, Tanzania, United States, and Vietnam
- Approved track enrollments: 19 in Track 1, and 18 in Track 2
- Test phase submissions: 7 in Track 1, and 10 in Track 2 (including baseline systems)
- System description reports for both tracks
Evaluation Statistics
- 65 listening tests conducted
- 17 codecs evaluated at two bitrate modes on specialized test sets
- Approximately 186,720 individual ratings of audio files gathered
- 24,140 crowdsourced participants (not all unique) contributed to the evaluation
System Description Reports
- Combined Reports (50 pages): All Technical Reports Combined
- Individual Reports: can be found in specific track section below
What’s Next
- Results analysis: Detailed analysis of the challenge results will be presented in subsequent work. We will keep participants informed.
- Workshop archival paper deadline: October 22, 2025 (further details: link)
- Workshop date and location: May 4, 2026 in Barcelona, Spain (further details: link)
We look forward to seeing you at the workshop and discussing the results of your work!
Track 1 — Transparency Codecs
Top-Ranked Entries
Top three submissions:
- Winners: teamwzqaq team (report)
- Runner up: nano_codec team (report)
- Third place: atid_go team (report)
Congratulation to the winning team, teamwzqaq, and to all the teams that participated in Track 1 of the 2025 LRAC Challenge.
Results Table
Figure 1: Track 1 - Transparency Codecs - Results Table
Leaderboard
- Track 1 Leaderboard (sortable results): link
System Reports
Individual Team Reports:
-
teamwzqaq team — Report
LRAC System Description for Track1 and Track2
Ziqian Wu, JiaWei Jiang, Kunpeng Lin, He Wang, Qingbo Huang
ByteDance -
nanocodec team — Report
NanoCodec: Towards Low Bitrate and Low Complexity Real-Time Neural Audio Codec
Andong Li, Pinglin Xu, Zhe Han, Lingling Dai, Yiqing Guo, Hua Gao, Xiaodong Li, Chengshi Zheng
Institute of Acoustics, Chinese Academy of Sciences, Beijing, China
ByteDance, China
University of Chinese Academy of Sciences, Beijing, China -
atid_go team — Report
Low Resource Audio Codec Challenge Track1: Transparency Codec
Zixiang Wan, Guochang Zhang, Haoran Zhao, Runqiang Han, Jianqiang Wei
Anker Innovations, Beijing, China -
nju-aalab team — Report
VoCodec: An Efficient Lightweight Low-Bitrates Speech Codec
Leyan Yang, Ronghui Hu, Yang Xu, Jing Lu
Key Laboratory of Modern Acoustics, Nanjing University, Nanjing 210093, Jiangsu, China
NJU-Horizon Intelligent Audio Lab, Horizon Robotics, Beijing 100094, China -
boya_audio team — Report
HorCodec: Hornet Based Neural Audio Codec for the LRAC 2025 Challenge Track 1
Qingbo Huang, Weihao Xiong, Congxin Zhang, Xinmin Yan
ByteDance -
pdura7 team — Report
Low Resource Audio Codec Challenge Sublime System Description
Piotr Dura
Sanas.ai -
Baseline Systems — Report
Baseline Systems for the 2025 Low-Resource Audio Codec Challenge
Yusuf Ziya Isik, Rafał Łaganowski
Collaboration AI, Cisco Systems, Inc.
Track 2 — Speech Enhancement Codecs
Top-Ranked Entries
Top three submissions:
- Winners: nju-aalab team (report)
- Runner up: xuyang team (report)
- Third place: nano-codec team (report)
Congratulation to the winning team, nju-aalab, and to all the teams that participated in Track 2 of the 2025 LRAC Challenge.
Results Table
Figure 3: Track 2 - Speech Enhancement Codecs - Results Table
Leaderboard
- Track 2 Leaderboard (sortable results): link
System Reports
Individual Team Reports:
-
nju-aalab team — Report
Progressive Refinement Training for Low-Resource Neural Speech Coding and Enhancement
Ronghui Hu, Leyan Yang, Yang Xu, Qinwen Hu, Jing Lu
Key Laboratory of Modern Acoustics, Nanjing University, Nanjing 210093, Jiangsu, China
NJU-Horizon Intelligent Audio Lab, Horizon Robotics, Beijing 100094, China -
xuyang team — Report
A Low-Latency VQ-GAN-Based Codec with Knowledge Distillation for Joint Speech Coding and Enhancement
Yang Xu, Ronghui Hu, Leyan Yang, Jing Lu
Key Laboratory of Modern Acoustics, Nanjing University, Nanjing, 210093, Jiangsu, China
NJU-Horizon Intelligent Audio Lab, Horizon Robotics, Beijing, 100094, Beijing, China -
nanocodec team — Report
Enhance-NanoCodec: Enhancement Neural Audio Codec for the LRAC 2025 Challenge Track 2
Andong Li, Linping Xu, Zhe Han, Lingling Dai, Yiqing Guo, Hua Gao, Xiaodong Li, Chengshi Zheng
Institute of Acoustics, Chinese Academy of Sciences, Beijing, China
ByteDance, China
University of Chinese Academy of Sciences, Beijing, China -
atid_go team — Report
Low Resource Audio Codec Challenge Track2: Denoising Codec
Haoran Zhao, Zixiang Wan, Guochang Zhang, Runqiang Han, Jianqiang Wei
Anker Innovations, Beijing, China -
teamwzqaq team — Report
LRAC System Description for Track1 and Track2
Ziqian Wu, JiaWei Jiang, Kunpeng Lin, He Wang, Qingbo Huang
ByteDance -
boya_audio team — Report
Efficient Real-Time Audio Codec with Integrated Speech Enhancement Techniques
Weihao Xiong, Congxin Zhang, Xinming Yan, Qingbo Huang
ByteDance -
leyan team — Report
VoCodec: An Efficient Lightweight Low-Bitrates Speech Codec
Leyan Yang, Ronghui Hu, Yang Xu, Jing Lu
Key Laboratory of Modern Acoustics, Nanjing University, Nanjing 210093, Jiangsu, China
NJU-Horizon Intelligent Audio Lab, Horizon Robotics, Beijing 100094, China -
parslog team — Report
Low-Complexity End-to-End Speech Enhancement Codec for Real-Time Communication in Noisy and Reverberant Conditions
Pincheng Lu, Peng Zhou, Xiaojiao Chen, Jing Wang
Beijing Institute of Technology, Beijing, China -
pdura7 team — Report
Low Resource Audio Codec Challenge Sublime System Description
Piotr Dura
Sanas.ai -
Baseline Systems — Report
Baseline Systems for the 2025 Low-Resource Audio Codec Challenge
Yusuf Ziya Isik, Rafał Łaganowski
Collaboration AI, Cisco Systems, Inc.