PeopleofSuperintelligence

Researcher

Gary Wang

Gary Wang is a Google researcher whose work spans speech recognition and text-to-speech. He coauthored speech-augmentation and semi-supervised recognition methods, then contributed to research on accented speech, multilingual recognition and synthesis, and text injection for ASR. His 2025 work includes a robust ASR paper and a speech-text prompting patent application; a 2026 patent application covers zero-shot cross-lingual voice transfer for text-to-speech. As of October 2026, he works on speech recognition and text-to-speech at Google Research.

Now at
Google
Works
130
Badges

Publication numbers from OpenAlex, as of 22 October 2026. Citations and h-index are hidden because this OpenAlex record may be incomplete or mix in other authors.

Timeline

  1. 2020

    SCADA for speech recognition

    In 2020, he coauthors SCADA, which combines augmentation policies, consistency measures and model regularization to improve speech recognition. It reports an 18% relative reduction in word error rate on LibriSpeech.

  2. 2021

    Semi-supervised ASR

    In 2021, he coauthors a study that combines multilingual speech synthesis on unspoken text with Noisy Student Training on untranscribed audio. The methods yield up to a 14.4% relative reduction in word error rate on a voice-search task.

  3. 2022

    Accented speech recognition

    In 2022, he coauthors a study on accented speech recognition that discusses benchmarks, wav2vec 2.0 pre-training, training approaches and corpora for evaluating linguistic diversity.

  4. 2023

    Google USM

    The Google USM paper names him as a coauthor in 2023. It describes a single large ASR model for 100+ languages, pretrained on 12 million hours of unlabeled multilingual data spanning more than 300 languages.

  5. 2023

    Multilingual speech synthesis

    In 2023, he coauthors Virtuoso, a speech-text framework for multilingual text-to-speech. Models trained with Virtuoso produce more natural and intelligible speech than baseline models in seen languages, and synthesize speech for unseen languages without paired TTS data.

  6. 2024

    Text injection with ASTRA

    In 2024, he coauthors ASTRA, which injects text into ASR training and aligns speech and text through a weighted RNNT objective. The method matches duration-based approaches on the FLEURS benchmark.

  7. 2025

    Robust ASR and speech prompting

    In 2025, he coauthors Speech Re-painting for Robust ASR with Kyle Kastner and colleagues. A Google patent application published in September 2025 names him as an inventor on Speech-text prompting for speech tasks; the method resynthesizes speech in a reference speaker’s voice.

  8. As of October 2026

    Google speech research

    As of October 2026, he works on speech recognition and text-to-speech at Google Research. A patent application published in March 2026 names him as an inventor on a zero-shot cross-lingual voice-transfer method for text-to-speech.

  1. 16% of AI career moves in the 2020s went from a university to a companyIncludes Gary Wang. We estimate that 16% of AI career moves in the 2020s went from a university to a company; 29% of moves starting at a university went to a company, and 76% ended the university role.
  2. Google had the most people shaping AI every year from 2014 to 2025Includes Google. Our estimates of which organizations employed the most people shaping AI each year from 1955 to 2025, tracing the shift from university labs to companies.
  3. Google has the highest combined h-index among AI organizationsIncludes Google. We estimate that Google holds the highest combined h-index at 8,016, while universities account for 73% of the deduplicated total across AI organizations.
  4. 20% of people at companies who shape AI also hold a university role, led by Google and Stanford UniversityIncludes Google. We estimate that 539 of the 6,766 people with a known current organization (8%) hold both a university role and a company role, including 20% of people at companies and 13% of people at universities.

Sources and credits

Photo: We have no published photo of Gary Wang yet. Send us a free-licensed photo.

  1. SCADA: Stochastic, Consistent and Adversarial Data Augmentation to Improve ASR research.google. Checked 22 October 2026.
  2. Semi-Supervision in ASR: Sequential Mixmatch and Factorized TTS-Based Augmentation research.google. Checked 22 October 2026.
  3. Accented Speech Recognition: Benchmarking, Pre-training, and Diverse Data research.google. Checked 22 October 2026.
  4. Google USM: Scaling Automatic Speech Recognition Beyond 100 Languages arxiv.org. Checked 22 October 2026.
  5. Virtuoso: Massive Multilingual Speech-Text Joint Semi-Supervised Learning for Text-to-Speech research.google. Checked 22 October 2026.
  6. ASTRA: Aligning Speech and Text Representations for Asr without Sampling arxiv.org. Checked 22 October 2026.
  7. Google at ICASSP 2025 research.google. Checked 22 October 2026.
  8. Speech-text prompting for speech tasks patents.google.com. Checked 22 October 2026.
  9. Gary Wang research.google. Checked 22 October 2026.
  10. Zero-Shot Cross-Lingual Voice Transfer for Text-To-Speech patents.google.com. Checked 22 October 2026.