April 2026
Research IT Solutions designed a new AI Transcription Service to meet an increasing need for a convenient and approved transcription solution for investigators.
Transcription of research participants’ audio and video recordings is often a critical component of qualitative research workflows across many fields of inquiry. Traditional human-mediated transcription processes can be costly, and novel AI-based automated methods require careful attention as recordings can contain sensitive information.
This centralized user-friendly service, developed in partnership with Hopkins’ IRBs, gives investigators an efficient, affordable, and compliant means to generate accurate transcriptions of recordings with diarization — the process of partitioning an audio recording based on the identity of each speaker.
The solution leverages Microsoft’s Batch Speech-to-Text (STT) service. It is wrapped into a Python-based Azure Function App deployed on Hopkins’ managed Research IT cloud infrastructure and configured to meet security and compliance requirements for personally identifiable information (PII)/protected health information (PHI).
To use the service, each research group is provisioned with a dedicated Azure Blob Storage container, accessible only via JHED-authenticated login, to upload their audio or video files and automatically trigger transcription. Within minutes, text-based transcriptions become available to investigators, who can transfer them to their research data storage folders.
A two-month-long pilot was conducted with a small number of research groups to gather user feedback and validate performance and scalability. Service approval steps included solution architecture and IT risk reviews, and consultations with the JHM AI and Data Trust, the IRB AI Consultation team, the Software Intake Process team, and the finance team to ensure the service meets all IT security, compliance, and governance requirements.
Broad rollout across the Johns Hopkins research community is slated for late Spring 2026.
