From Voice to Insight
Every agency and enterprise deserves its own AI office brain
VoiVision AI, co-founded with NEU NLP Lab, focuses on two core product lines — AI Meeting Secretary and AI Meeting Translation — delivering a four-in-one enterprise platform: in-house hardware × edge compute × AI engine × full OA integration.
Four-in-One Tech Stack
In-house HW
From pocket recorders to enterprise servers — full coverage.
Edge Compute
Embedded NPU heterogeneous compute at 46–326 TOPS on NVIDIA / Ascend / Cambricon.
AI Engine
52 yrs of NEU NLP heritage since 1973: frontend + high-accuracy ASR.
Full OA
Deep integration with WeCom, DingTalk, Feishu, Landray.
ASR × OCR Multimodal Fusion
Beyond listening. VoiVision links on-device OCR with ASR, using the screen (PPT / docs / whiteboard) to boost speech recognition — BJ66+WCB06 dual-channel capture, on-device large-model OCR frame-by-frame, 200+ discipline KBs injecting hotwords into the decoder instantly.
Explore the tech →Two Core Product Lines
AI Meeting Secretary + AI Meeting Translation, powered by one AI engine
V06 Voice Card
Card-sized · Carry anywhere
The V06 is a business-card-sized portable recorder with a CNC aluminum body, dual omni mics plus bone conduction. One-tap recording and automatic sync to cloud or on-prem server turn personal meetings into searchable knowledge.
Learn more →AI Meeting Secretary
VV01 Meeting Host
1-channel · offline transcription
The VV01 is a desktop offline meeting host with 46 TOPS of edge compute. Plug in an external omnidirectional mic for pure on-device transcription under 1s latency at ≥98% accuracy — with speaker/voiceprint separation and auto minutes, ideal for small rooms and desks.
Learn more →AI Meeting Secretary
VV05 Meeting Server
5 channels · edge NPU
The VV05 is a department-grade 5-channel meeting server with an edge NPU at 166 TOPS and 24G high-bandwidth memory. Full local processing, zero data egress, distributed audio aggregation and a unified web console — built for R&D centers, financial institutions, and chain hubs.
Learn more →AI Meeting Secretary
VV10 Enterprise Server
10 channels · full private deployment
The VV10 is an enterprise 10-channel engine with an edge NPU at 326 TOPS and 48G high-bandwidth memory. Full local processing, zero data egress, unified web console — meeting MLPS Level 3 and classified requirements for government, defense, finance, and healthcare.
Learn more →AI Meeting Translation
VT01 AI Meeting Translator
1-channel bidirectional · live interpretation
The VT01 is a desktop AI meeting translator with 46 TOPS edge compute and a translation LLM. It supports 1-channel bidirectional live interpretation, bilingual subtitles, and TTS output — plug-and-play and fully offline for cross-border negotiations, international video calls, and multilingual stores.
Learn more →AI Meeting Translation
VT05 AI Translation Server
5-channel dual-screen · enterprise interpretation
The VT05 is a department-grade 5-channel dual-screen translation server with an edge NPU at 166 TOPS and 24G VRAM. It supports 5 concurrent bidirectional interpretation lanes, distributed audio aggregation, and a unified web console — for global R&D centers, cross-border business units, and multinational retail chains.
Learn more →AI Meeting Translation
VTD15 Desktop Translation Terminal
15" dual touchscreens · interpreter / guest view
The VTD15 is a desktop dual-screen translation terminal paired with VT01 / VT05 hosts. It features dual independent 15-inch HD touchscreens, directional array-mic far-field pickup, separate interpreter and guest displays, and real-time bilingual subtitles — plug-and-play for cross-border negotiations and international receptions.
Learn more →Industry Solutions
Government, finance, defense, healthcare & more
Government Meetings
Intranet · confidential
For government intranet meetings, delivers fully local, air-gapped minutes capability meeting MLPS Level 3 and classified requirements — zero data egress.
View solution →Defense & R&D
Air-gapped · zero link
Provides air-gapped, zero-link meeting intelligence for defense and research institutes, securing core R&D discussions while meeting high-concurrency needs.
View solution →Financial Institutions
Compliant archive · MLPS L3
For banks, securities, and insurers — compliant archiving and MLPS L3-ready meetings where decisions are traceable and auditable.
View solution →Healthcare
Consults · privacy compliant
Localized minutes for hospital consults and case discussions, protecting patient privacy and meeting healthcare data compliance.
View solution →Enterprise R&D
Parallel teams · no Queue
Department-grade concurrent minutes for large enterprise R&D — parallel team meetings with no queue, auto-crystallized into enterprise knowledge.
View solution →Retail Chains / Hubs
Central mgmt · aggregated
Unified management and aggregated minutes for retail chains and regional hubs — local store transcription, HQ-wide search, fast ops loop.
View solution →Blog & Insights
Deep dives on tech, industry and products
Beyond 98% Accuracy: The Five Engineering Stages That Decide Meeting Minutes Quality
Transcription accuracy is only the entry ticket. Whether minutes are actually usable depends on five more stages: speaker separation, summary templates, action-item extraction, passage polishing, and a genuinely offline chain. This article walks through each stage from a buyer's acceptance-testing perspective, with the common failure modes and a checklist you can use directly.
Read more →
Running meeting transcription on Ascend 310P: version matrix, model conversion and seven field lessons
Can an edge inference card handle meeting transcription? This article documents deploying a Chinese ASR engine on the Ascend 310P — CANN environment alignment, the ATC parameters that matter when converting ONNX to OM, the four devices a container must mount, how to set dynamic dimensions, and seven lessons you only learn on site. The most dangerous one is a silent operator failure that produces all-zero output without any error.
Read more →
Running Multilingual Meetings Offline: On-Prem Translation and Subtitle Sync
In a four-party meeting spanning Chinese, English, Japanese and Korean, the hard part is not recognition — it is knowing who said what, what it should become in another language, whether it is real time, and whether the audio may leave the building. This article breaks down the full multilingual meeting pipeline: language detection, time alignment between streaming ASR and translation, subtitle sync, terminology consistency, and why cloud interpretation rarely clears compliance in government and enterprise settings.
Read more →Book a Personalized Demo
Tell us your meeting scenario and compliance needs — get a tailored plan.