Speech recognition, real-time translation and AI minutes form three engines covering the core meeting-record workflow.
Speech recognition engine
ASRUnderstands dialects, terminology and fast on-screen text
Industry hotword customisation for greater accuracy in professional scenarios.
2–16 concurrent audio inputs with real-time on-screen text.
Offline transcription: 1 GB of audio in 5 minutes or less.
Mandarin, common dialects and multilingual recognition.
Real-time translation engine
MTMultilingual translation · bilingual on screen
Custom terminology libraries keep translations aligned with the industry.
Flagship edition transcription latency of 0.5 seconds or less.
Suited to international negotiations and conferences.
Real-time multilingual translation with bilingual subtitles on screen.
AI minutes engine
LLMUnderstanding and summarisation · not template stitching
Extract key conclusions, resolutions and open questions.
Extract action items with owners and deadlines.
Multidimensional summaries by topic or speaker.
Intelligent summarisation by topic, not template stitching.
Above the engines: four platform foundations.
02 CORE CAPABILITIES
Four platform foundations
Make recognition, translation and minutes work reliably in real meeting environments.
Voiceprint speaker library
Speaker registration, voiceprint modelling and automatic attribution.
Industry lexicon management
Custom hotwords and terminology to continuously improve recognition accuracy.
Security and permissions
On-premises deployment, tiered authorisation and audit trails.
Open integration
API integration with mainstream meeting systems and write-back to OA / ticketing.
ConnectivityVenue audio capture
TranscriptionReal-time speech on screen
TranslationBilingual subtitle output
MinutesTopic summarisation
Closed loopArchiving and write-back of action items
03 DEPLOYMENT & HARDWARE
Four device categories for a complete operating environment
Software needs a complete hardware foundation: platform, compute, venue terminals and audio capture must all be present for real-time transcription and minutes to work.
Application and database servers
Required
Handles meeting management, user permissions, transcription scheduling, translation result processing, meeting minutes and data reports.
AI inference server
Required
Runs model inference for speech recognition, real-time transcription and translation; supports about 10 concurrent meetings (4 hours each).
Venue interaction unit
Configured per scenario
Connects venue audio devices, AI inference and platform services for real-time transcription, bilingual subtitles and minutes write-back.
USB audio capture unit / networked audio capture terminal
Configured per scenario
Connects microphones, mixers and audio matrices, feeding analogue or digital venue audio to the AI server for real-time transcription and translation.
The platform server manages business and data; the AI inference server determines concurrency and latency; the venue interaction unit and audio capture devices determine what the venue actually hears and sees.
04 WORKFLOW
How the system connects
Venue audio enters the AI inference server through the capture terminal for transcription and translation; results are written to the platform server, while users view minutes and subtitles appear on screen through the venue interaction unit.
01
Venue audio devices
Microphones / mixer / audio matrix.
02
Audio capture terminal
USB capture unit / networked audio capture terminal.