AI Extensions
AI built into every call and meeting
A suite of AI features on the SRTC engine: live transcription, AI minutes, face verification and voice masking, all deployable on your own servers.
AI Extensions are a set of AI features for SRTC and STMLink Meeting: live transcription, AI meeting minutes, face verification at join, and voice masking. Transcription runs at 92%+ accuracy with millisecond-level latency, separates speakers and handles English, Chinese and mixed speech; minutes are generated by an LLM with keywords, chapters and summaries at several levels of detail. Every feature can be self-hosted, so audio, video and text stay on your network.
At a glance
| Item | Details |
|---|---|
| Transcription | Live and post-call, 92%+ accuracy, millisecond-level latency, speaker separation, English / Chinese / mixed |
| Meeting minutes | Keywords, chapters, summaries at three levels of detail |
| Face verification | At join and continuous, with alerts on mismatch |
| Voice masking | Real-time, multiple voice styles |
| Works with | SRTC Engine, STMLink Meeting |
| Deployment | Self-hosted, with your own ASR / LLM if you prefer |
Live transcription
Speech becomes text during the meeting, split by speaker and sentence
- Live transcription and transcription of recordings, with millisecond-level latency
- Speaker separation: each participant’s lines shown separately
- Sentence segmentation from pauses, semantics and prosody
- English, Chinese and mixed speech
- 92%+ accuracy on live transcription
- Live captions, live transcript and full history
Self-hosted: data stays on your network
AI meeting minutes
LLM-generated minutes that pull out what matters
- Keywords, with transcript search by keyword
- Automatic chapters for each phase of the meeting
- Short, medium or detailed summaries
- Find everything a given speaker said
Self-hosted: data stays on your network
Face verification
Verify that the right people are in the meeting
- Only participants who pass face verification can join
- Continuous verification for designated members
- Alerts when verification fails
- For high-security government and enterprise meetings
Self-hosted: data stays on your network
Voice masking
Real-time voice changing to protect speakers
- Real-time processing without noticeable delay
- Multiple voice styles
- For anonymous sessions and privacy-sensitive meetings
Self-hosted: data stays on your network
Ready to build?
Get a self-hosted trial, SDK access and hands-on help from our engineers