This new availability builds on the recent release of Bridge2AI Voice v2.0, which already includes more than 16,000 recordings from 442 participants across five North American sites. Previous releases provided derived features such as spectrograms, acoustic measures, and clinical data, but not the raw recordings themselves. Making the original audio available under strict privacy protections will allow researchers to pursue deeper analyses while ensuring participant safeguards remain in place.

“Access to raw audio enables researchers to explore new acoustic and computational features that derived data alone may not capture,” said Yael Bensoussan, MD, Principal Investigator of Bridge2AI Voice. “This release will accelerate the science of voice biomarkers while maintaining the highest standards of privacy and ethics.”

Funded by the NIH Common Fund, Bridge2AI Voice is creating a diverse, ethically sourced dataset of 10,000 voices linked to multimodal health information. By sharing these resources responsibly, the project is fueling innovation in artificial intelligence and supporting the integration of voice as a biomarker of health in clinical care.

Bridge2AI is a broader NIH Common Fund program designed to propel the field of artificial intelligence in biomedicine by creating ethically sourced, high-quality datasets and developing best practices for their use. Through collaborations spanning biomedical research, computer science, ethics, and social sciences, Bridge2AI is helping build the infrastructure needed to ensure AI tools are accurate, equitable, and impactful for diverse populations.