Interspeech2025 URGENT Challenge
Summary: The URGENT challenge series aims to foster the development of Universal, Robust, and Generalizable speech EnhancemeNT systems. The Interspeech 2025 URGENT challenge is the second edition in the series, which focuses on (i) investigating how to leverage noisy corpora effectively (ii) addressing a wide range of speech degradations, and (iii) exploring language dependency of speech enhancement models. We welcome both discriminative and generative models. The challenge has two tracks with different training data scales (~2.5k and ~60k hours), allowing participants to investigate the scalability of the system. Evaluation will be done using as many as 13 objective metrics and subjective listening.
Organizers: Kohei Saijo, Wangyou Zhang, Samuele Cornell, Robin Scheibler, Chenda Li, Zhaoheng Ni, Anurag Kumar, Marvin Sach, Yihui Fu, Wei Wang, Tim Fingscheidt, Shinji Watanabe
Challenge website: https://urgent-challenge.github.io/urgent2025/
MultiLingual Speech processing Universal PERformance Benchmark (SUPERB) Challenge
Summary: Multilingual SUPERB (ML-SUPERB) is an extension of the SUPERB benchmark, designed to evaluate the cross-lingual capabilities of speech representation learning. For this yearâs challenge, we invite participants to develop state-of-the-art ASR systems for all languages and language varieties. As such, the ML-SUPERB 2.0 challenge is fully unconstrained: we encourage the use of the latest developments in foundation models and data curation. The challenge will feature a live leaderboard and online evaluation server that tests the robustness of model submissions across 154 languages and over 200 accents and dialects.
Organizers: Antonios Anastasopoulos, Martijn Bartelds, William Chen, Dan Jurafsky, Hung-yi Lee, Karen Livescu, Chutong Meng, Jiatong Shi, Hsiu-Hsuan Wang, Shih-Heng Wang, Shinji Watanabe
Challenge website: https://multilingual.superbbenchmark.org
Multimodal Information Based Speech Processing (MISP) 2025 Challenge
Summary: Meetings are one of the most valuable yet challenging scenarios for speech applications, as they are rich in information exchange and decision-making processes, making accurate transcription and analysis of the content crucial for enhancing productivity and preserving insights. The MISP 2025 challenge focuses on multi-modal multi-device meeting transcription and aims to push the boundaries of current techniques by introducing additional modality information, specifically the video modality. The specific tasks considered in the challenge are: 1. Audio-Visual Speaker Diarization, 2. Audio-Visual Speech Recognition, and 3. Audio-Visual Diarization and Recognition.
Organizers: Hang Chen, Jun Du, Chin-Hui Lee, Sabato Marco Siniscalchi, Shinji Watanabe, Jingdong Chen, Odette Scharenborg
Challenge website: https://mispchallenge.github.io/mispchallenge2025/index.html
Speech Accessibility Project Challenge
Summary: The goal of the Speech Accessibility Project Challenge is to rapidly advance the state of the art in dysarthric speech recognition. Competitors can use data from the Speech Accessibility Project 2024-04-30 dataset (train: 290 hours from 369 speakers, dev: 44 hours from 59 speakers) to train an ASR. Winners will include the system with the lowest word error rate on the test set, and the team whose transcripts have the highest semantic score.
Organizers: Mark Hasegawa-Johnson, Aadhrik Khuila, Alicia Martin, Brian Gamido, Christopher Zwilling, Colin Lea, Ed Cutrell, Gautam Mantena, Katrin Tomanek, Kyu Jeong Han, Leda Sarı, Venkatesh Ravichandran
Challenge website: https://eval.ai/web/challenges/challenge-page/2362/overview
Speech Emotion Recognition in Naturalistic Conditions Challenge
Summary: The Speech Emotion Recognition (SER) in Naturalistic Conditions Challenge at Interspeech 2025 aims to advance the field of emotion recognition from spontaneous speech, emphasizing real-world applicability over-controlled, acted scenarios. Utilizing the MSP-Podcast corpus â a rich dataset of over 324 hours of naturalistic conversational speech â the challenge provides a platform for researchers to develop and benchmark SER technologies that perform effectively in complex, real-world environments. Participants have access to speaker-independent training and development sets, as well as an exclusive test set, all annotated for two distinct tasks: categorical emotion recognition and emotional attributes prediction.
Organizers: Carlos Busso, Berrak Sisman, Najim Dehak, Lucas Goncalves, Abinay Reddy Naini, Ali N. Salman, Pravin Mote, Thomas Thebaud, Laureano Moro-Velazquez, Leibny Paola Garcia Perera
Challenge website: https://lab-msp.com/MSP-Podcast_Competition/IS2025/
The 1st Mice Autism Detection via Ultrasound Vocalisation (MAD-UV) Challenge
Summary: The 1st Mice Autism Detection via Ultrasound Vocalisation (MAD-UV) Challenge invites participants to develop models to classify mice as either wild-type or modelling Autism Spectrum Disorder (characteristics). Participants are encouraged to use cutting-edge speech and signal processing techniques to explore high-frequency, high-sampling-rate audio data, with the potential to explore the possibility that individual differences in these representations from mice models are parallel to those found in human speech research.
Organizers: Yoshiharu Yamamoto, Björn W. Schuller, Toru Takumi, Zijiang Yang, Lukas Christ, Meishu Song
Challenge website: https://www.mad-uv.org/
The 1st SpeechWellness Challenge
Summary: Suicide is a significant global health problem, especially for young people. The SpeechWellness Challenge aims to detect suicide risk in teenagers through speech analysis. Using both spontaneous and reading speech data from 600 adolescents, including those identified as at risk, participants aim to develop models to identify digital biomarkers associated with suicidal tendencies. The challenge encourages diverse approaches, including signal processing, speech and emotion recognition, and large language models. By using speech data, the challenge aspires to advance the early detection of mental health issues and provide scalable, accessible tools for intervention in adolescent suicide prevention.
Organizers: Chao Zhang, Runsen Chen, Ziyun Cui, Chang Lei, Wen Wu, Diyang Qu, Yinan Duan, Ji Wu
Challenge website: https://speechwellness.github.io/
Interspeech 2025
PCO: TU Delft Events
Delft University of Technology
Communication Department
Prometheusplein 1
2628 ZC Delft
The Netherlands
Email: pco@interspeech2025.org
X (formerly Twitter): @ISCAInterspeech
Interspeech 2025 is working under the privacy policy of TU Delft