WAVLab | @CarnegieMellon

Shinji Watanabe's Audio and Voice Lab | WAVLab @LTIatCMU @SCSatCMU | Speech Recognition, Speech Enhancement, Spoken Language Understanding, and more.

[Shinji Watanabe @shinjiw_at_cmu]

My Google Scholar h-index just reached 100! 🙂

My first paper was in 1999, and behind these 100 are students and collaborators in academia and industry, plus the open-source community and challenges.

This number is really yours. Thank you all! 🙏

https://t.co/IVGQM30qe8

[William Chen @chenwanch1]

Who wants more speech training data?

YODAS v3 is now available @huggingface

It’s 1.1M hours - the biggest audio dataset ever. And the first at this scale with stereo audio at 48kHz, with timestamped transcripts and translations. 100+ langs

https://t.co/KLS44NvDC5

CC-BY-3.0

[WAVLab | @CarnegieMellon @WavLab]

Next up, join our afternoon tutorial:

Sun, Sep 27 · 13:30-17:00 · Room C3.3
Conversational Speech Recognition and Analysis - Progress, Challenges, and Emerging Opportunities with Speech Language Models

[WAVLab | @CarnegieMellon @WavLab]

WAVLab is at #Interspeech2026 in Sydney 🇦🇺

Our papers cover speech recognition, spoken LLMs, speech synthesis, speech quality, and multilingual data. Plus the new ESPnet3 toolkit!

Here's where to find us, Sep 28 to Oct 1 🧵

[WAVLab | @CarnegieMellon @WavLab]

Thanks to everyone who came to our morning tutorial, "On the Landscape of Spoken Language Models"!

[WAVLab | @CarnegieMellon @WavLab]

Come say hi! All our papers: https://t.co/IRFobFjCM5

#Interspeech2026

[WAVLab | @CarnegieMellon @WavLab]

Thu, Oct 1

15:20 · Room C2.5
Endpoint Anticipation for Low-Latency Spoken Dialogue
https://t.co/VKgp5Lpyk1

[WAVLab | @CarnegieMellon @WavLab]

Thu, Oct 1

14:00-16:00 · Poster, Gallery p7
Dissecting Sensitivity to Training Language in Self-Supervised Speech Learning Using Neural Audio Codec Tokens
https://t.co/gehxgVSkmY

[WAVLab | @CarnegieMellon @WavLab]

Thu, Oct 1

09:00-11:00 · Room C3.6 (special session)
Adapting Text LLMs to Speech via Multimodal Depth Up-Scaling
https://t.co/oERBlWZVwc

10:40 · Room C3.3
Bagpiper-TTS: Natural Language Guided Universal Speech Synthesis
https://t.co/ndMAD62pmv

[WAVLab | @CarnegieMellon @WavLab]

Wed, Sep 30

17:50 · Room C2.1
URGENT-MOS: Unified Multi-Metric and Preference Learning for Robust Speech Quality Assessment
https://t.co/AGQhzpXAvP

[WAVLab | @CarnegieMellon @WavLab]

Wed, Sep 30

17:10 · Darling Harbour Theatre
Evaluating Large Language Models Abilities for Addressee, Turn-change, and Next Speaker Prediction in Meetings
https://t.co/2aDGsLrwlD

[WAVLab | @CarnegieMellon @WavLab]

Wed, Sep 30

16:50 · Room C2.1
ANCHOR: Autoregressive Non-intrusive Chunk-Ordered Refinement for Joint Multi-Resolution Speech Quality Modeling
https://t.co/l4cJJyUT0U

[WAVLab | @CarnegieMellon @WavLab]

Wed, Sep 30

15:00 · Room C3.3
ESPnet3: Infrastructure for Scalable Speech and Audio Research in the Foundation Model Era
https://t.co/LiXIuaYHTo

15:40 · Room C2.6
Bagpiper-Edit: Zero-Shot Open-Ended Audio Editing via Rich-Caption
https://t.co/gkL5LQNMrY

[WAVLab | @CarnegieMellon @WavLab]

Wed, Sep 30

14:00-16:00 · Poster, Gallery p6
YODAS v3: Over 1 Million Hours of High-Bandwidth, Stereophonic, Multilingual Speech
https://t.co/QtLp0OdTS2

[WAVLab | @CarnegieMellon @WavLab]

Wed, Sep 30

09:00-11:00 · Poster, Gallery p6
An Efficient vLLM-Based Inference Pipeline for Unified Audio Understanding and Generation
https://t.co/GQKFmFifej

[WAVLab | @CarnegieMellon @WavLab]

Wed, Sep 30

09:00 · Room C2.2
Comparative Reasoning: Making an Audio Language Model Better at Comparing Emotions
https://t.co/wGUiCesILA

[WAVLab | @CarnegieMellon @WavLab]

Tue, Sep 29

17:30 · Room C2.2
An Empirical Recipe for Universal Phone Recognition
https://t.co/t2V0aVWfET

[WAVLab | @CarnegieMellon @WavLab]

Mon, Sep 28

14:30-16:30 · Poster, Gallery p4
Online Predictive Coding for Dual-Mode Self-Supervised Speech Models
https://t.co/gAVQdEOl7n

[WAVLab | @CarnegieMellon @WavLab]

Mon, Sep 28

12:40 · Room C3.4
Who Spoke What When? Evaluating Spoken Language Models for Conversational ASR with Semantic and Overlap-Aware Metrics
https://t.co/QDvsGSLHPM

[WAVLab | @CarnegieMellon @WavLab]

Mon, Sep 28

11:40 · Darling Harbour Theatre
Which Data Matter? Embedding-Based Data Selection for Speech Recognition
https://t.co/JUYaILmODl

12:00 · Room C3.4
Speaker-Aware Hypothesis Clustering and Merging for Target-Speaker-free and Target-Speaker Multi-Talker ASR

[WAVLab | @CarnegieMellon @WavLab]

Mon, Sep 28

11:00 · Room C3.4
Mind the Gap: Impact of Synthetic Conversational Data on Multi-Talker ASR and Speaker Diarization
https://t.co/Dsfln8jwzX

11:20 · Room C3.4
Grounding Spoken LLMs in Multi-Speaker Audio via Diarization Conditioning
https://t.co/tBsOA3SZkG