Case Study: Zoom achieves sub-300ms real-time meeting transcription with Gladia

Zoom powers real-time meeting bots with Gladia in under 300ms

The customer zoom sought to build a reliable automated meeting assistant. The primary challenge was developing a transcription layer that could deliver sub-300ms latency, accurately handle multi-speaker conversations with real-time diarization, and support code-switching between languages, all while integrating with conferencing platform APIs like Zoom's Real-Time Media Streams. To power this, they used the real-time transcription API from the vendor Gladia.

Gladia provided a managed API solution that handled the complex infrastructure, including low-latency streaming, built-in speaker diarization, and support for over 100 languages with code-switching. This eliminated the need for zoom to manage costly GPU infrastructure and complicated integrations, allowing them to focus on their application's core features. The results were a reliable meeting bot with the required sub-300ms latency and high accuracy, even with accented speech and multilingual participants, enabling trustworthy automated summaries and action items.

zoom