Shenzhen Tendzone Intelligent Technology Co., Ltd. is one of the most reliable manufacturers and suppliers of ai large language model intelligent voice application host in China. Please feel free to wholesale high quality ai large language model intelligent voice application host made in China here from our factory. We also accept customized orders.
AI Voice Conference Host
Powered by Large Model Intelligence
Fully Offline Natural Language Control | Smart Transcription & Meeting Minutes
Built on a high-performance domestic AI acceleration chip and deeply integrated with large language models, the AI-100 brings audio-visual control together with intelligent voice collaboration to empower fully-scenarios conferences and deliver a new standard of intelligent communication.
Product Overview
The AI-100 AI Large Language Model Intelligent Voice Application Host is a fully offline intelligent speech transcription and meeting collaboration device designed for high-end scenarios including government, finance, education, judiciary, and enterprise meetings. Powered by a domestic high-performance ARM architecture and dedicated AI acceleration chips, it supports concurrent multi-channel audio processing, an AI voice assistant, real-time speech transcription, intelligent subtitle overlay, voiceprint recognition, Chinese-English mixed transcription, and automatic meeting minutes generation - all without internet access or additional servers, ensuring data security and system stability.
As the central intelligent voice node of the modern conference room, the AI-100 bridges microphones, conference systems, audio processors, mixers, network audio streams and video sources. It converts spoken conversations into structured, searchable, and shareable knowledge, enabling every meeting to become a continuously accumulating organizational asset.
Product Highlights
|
01 Natural Language Control Effortless and intuitive voice control experience. |
02 Real-time Speech-to-Text Precise voiceprint recognition and instant transcription. |
03 Smart Subtitle Overlay Real-time voice broadcasting on HDMI 4K video output. |
|
04 Intelligent Meeting Minutes Auto-generated structured meeting summaries. |
05 Secure & Reliable Fully offline operation, data never leaves the device. |
06 Powerful Performance Flexible expansion with 26 TOPS AI compute. |
AI+ Processing Capability
AI Speech Recognition
Professionally optimized models deliver ≥98% recognition accuracy for Mandarin, American English, and mixed-language speech. End-to-end streaming latency stays ≤100ms, with automatic punctuation, intelligent correction of homophones, and industry-specific term recognition.
AI Voiceprint & Speaker Diarization
Extracts speaker voiceprint features and distinguishes roles in real time. The built-in voiceprint library supports registration, renaming, deletion, and can be linked to microphones to auto-label speaker names in meeting minutes.
AI Intelligent Summarization
Built-in on-device large language model summarizes transcribed meeting content and audio/video files. Multiple templates - summary, conclusion, multi-party discussion, reporting - are selectable by meeting type to generate conclusions and action items with one click.
AI Voice Assistant
Voice wake-up and multi-turn dialogue enable device info queries, IoT and meeting-room device control, video conference call control, and rapid scene-preset dispatch - no precise commands required.
User Value
|
Save Time & Labor No dedicated secretary needed; meeting records are produced automatically. |
Multi-language Translation Real-time cross-language communication with Chinese-English mixed recognition. |
|
Multi-purpose Device One device covers transcription, subtitles, minutes, voice assistant, and more. |
Full-scenario Adaptation Strong system compatibility for conferences, training, command, and court scenarios. |
Application Scenarios
From government and enterprise boardrooms to courtrooms and command centers, the AI-100 adapts to the full spectrum of professional meeting environments.
|
Government Internal Meetings • Fully offline transcription • Speaker role separation • Sensitive-word filtering • Encrypted minutes export |
Enterprise & Cross-border Meetings • Multi-language real-time transcription • Voiceprint identification • Real-time translation • Intelligent minutes |
Education, Training & Court Records • High-accuracy transcription • Batch file transcription • VSD/CC subtitling • Simultaneous recording |
|
Remote Video Conferencing • HDMI subtitle overlay • Meeting summaries • Audio loop-out • 4K@60fps support |
Smart Courtrooms & Interrogation Rooms • Multi-channel recording • Real-time subtitles • Speaker role labeling • Voiceprint library management |
Command & Dispatch Centers • AI voice assistant • Business-system integration • Real-time on-screen subtitles • Device control |
Key Features
01 Secure & Reliable, Fully Offline
Built-in local LLM and speech recognition models run entirely on-device. No internet connection is required at any time, safeguarding meeting content and fully meeting government and enterprise data-security compliance requirements.
02 Accurate Recognition, Efficient Collaboration
Professionally optimized models reach ≥98% recognition accuracy for Chinese, English, and mixed-language speech. Combined with voiceprint recognition and speaker diarization, the AI-100 automatically identifies speakers and generates structured intelligent minutes, greatly improving meeting efficiency.
03 4K Smart Subtitles, Seamless Integration
Real-time overlay of customizable subtitles on HDMI 4K@60fps video - perfectly suited for video conferencing and local recording. The device supports analog audio, USB microphones, and network audio sources, seamlessly integrating with existing meeting systems.
04 Powerful Performance, Flexible Expansion
The dedicated NPU + GPU + coprocessor architecture delivers up to 26 TOPS of compute, supporting 16 concurrent PCM audio streams and 4 concurrent real-time transcription streams. An open API enables easy integration with paperless, control, and third-party business platforms.
05 Intelligent Post-Processing, Greater Value
Beyond transcription, the AI-100 supports intelligent summarization, colloquial text polishing, AI voice assistant (voice control), voice broadcast, and meeting audio recording, turning meeting records directly into valuable, reusable knowledge assets.
06 Unified Management, Easy O&M
A full-featured web-based configuration interface enables remote O&M, status monitoring, and data statistics. Centralized management simplifies large-scale deployment across multiple rooms and sites.
Device Specifications
Software Features
|
Item |
Specifications |
|
Offline Transcription |
Built-in ASR speech recognition - no internet or external server required. |
|
Multi-language Support |
Real-time transcription and mixed recognition of Mandarin Chinese and English. |
|
On-device LLM |
Dedicated compute module with a built-in locally quantized LLM; single-unit offline inference decode TPS ≥ 50 tokens/s, memory usage < 6GB; no internet required; data stays on-device. |
|
Audio Inputs |
Supports wired microphones, conference systems, audio processors, mixers, and USB omnidirectional microphones; supports RTSP and PCM stream audio. |
|
Audio Recording |
Records meeting audio and generates MP3 files. |
|
Transcription Performance |
Real-time streaming recognition with end-to-end transcription latency ≤ 100 ms for near-synchronous speech-to-text output. |
|
Transcription Accuracy |
≥98% real-time transcription accuracy for standard Mandarin or standard American English in clear, noise-free environments. |
|
Automatic Punctuation |
Predicts and automatically inserts Chinese punctuation - commas, periods, question marks, exclamation marks, enumeration commas, book title marks - in real time. |
|
Intelligent Correction |
Dedicated inference model with contextual semantic understanding provides powerful correction of homophones and similar-sounding words. |
|
Intelligent Segmentation |
Automatic segmentation based on semantic completeness, keywords, VAD combined with keywords or character count; users can customize pause duration and segment length thresholds. |
|
Keyword Filtering & Optimization |
Supports filler-word filtering, sensitive-word masking, and keyword replacement, plus recognition of industry-specific and professional terminology. Supports configuration and batch import of filler words, sensitive words, and keywords. |
|
Multi-channel Transcription |
High-performance NPU supports 16 concurrent real-time PCM audio streams and 4 concurrent real-time transcription streams. |
|
Real-time Subtitles |
Real-time synchronized subtitles overlay on HDMI input, RTSP streams, and local video files. Supports subtitle overlay on 4K@60fps HDMI input with loop-out of subtitled video. Flexible settings for font, size, color, background, transparency, and on-screen position. |
|
Voiceprint Recognition |
Extracts and analyzes speaker voiceprint features, distinguishing speakers for role separation. Provides voiceprint library management with registration, renaming, and deletion; can associate microphones or speaker info to auto-label speaker names in minutes. |
|
File Transcription |
Supports offline transcription of uploaded audio (MP3, WAV, AAC) and video (MP4, MOV, MKV) files; supports batch import with sequential transcription; file transcription efficiency up to 10:1 (10 minutes of audio/video transcribed in ~1 minute). |
|
Web Management |
Built-in web interface for transcription operations and configuration management. |
|
On-device Computing |
Built-in LLM supports text polishing, intelligent summarization, AI assistant, and more without network or compute servers. |
|
Text Polishing |
Automatically identifies colloquial or error-prone text issues - redundancy, disordered word order, inappropriate wording, missing logical connections, slips of the tongue - and corrects them by restructuring sentences, substituting context-appropriate words, and adding logical connectors, while preserving key information and tone. |
|
Intelligent Summarization |
Summarizes real-time transcribed meeting records and audio/video files. Multiple templates (summary, conclusion, multi-party discussion, reporting) can be selected by meeting type to generate conclusions and action items. |
|
AI Voice Assistant |
Voice wake-up and voice dialogue enable device info queries, IoT and meeting-room device control, video conference call control, and rapid scene-preset dispatch; supports multi-turn dialogue and intent recognition. |
|
System Integration & API |
Provides API interfaces for integration with paperless, control, hyper-converged, distributed, and third-party business platforms, enabling transcription control, real-time transcription, and file download within third-party systems. |
Hardware Specifications
|
Item |
Specifications |
|
Audio Input |
2× 3.5 mm audio input jacks |
|
Audio Output |
2× 3.5 mm audio output jacks |
|
HDMI Input |
1× HDMI HD input |
|
HDMI Output |
1× HDMI HD output (loop-out with subtitle overlay) |
|
Serial Ports |
2× RS-232, 2× RS-485/RS-422 |
|
USB |
1× USB 2.0, 1× USB 3.0, 1× USB Type-C |
|
LAN |
2× RJ45 Gigabit Ethernet ports |
|
Power Supply |
12V DC power |
|
Mounting |
1U telecom-standard rack-mount; install near peripheral devices to reduce cabling |
|
Dimensions |
Product: 436.8 × 45 × 318 mm (W × H × D); Package: 560 × 135 × 408 mm (W × H × D) |
|
Weight |
Net weight 3.3 kg; gross weight 4.75 kg (incl. packaging) |
|
Input Voltage / Current |
12V 5A |
|
Power Consumption |
40 W |
|
Operating Temperature |
0–45 °C |
|
Humidity & Altitude |
10%–90% RH; altitude ≤ 5,000 m |
Rear Panel Diagram
Our Service
|
Pre-Sales We assess your unique requirements, provide tailored consultations and demonstrations, and design a customized AV system to ensure the perfect solution for your scenario.
|
In-Sales We deliver transparent proposals, manage the full project implementation, and provide initial training to ensure a smooth and successful setup. |
After-Sales We offer prompt technical support, customized maintenance plans, continuous software upgrades, and actively use your feedback to drive ongoing service improvement. |
About Tendzone
Established in 2010, Tendzone is a global leader in providing advanced audio-visual (AV) solutions and manufacturing high-quality AV products. We specialize in a wide range of cutting-edge technologies, including audio processors, microphones, speakers, power amplifiers, AV over IP systems, digital conference systems, and MIDIS Distributed Multimedia Transmission Control Systems. Our solutions are trusted across industries such as conference rooms, command centers, education, multi-functional halls, and stadiums.
Hot Tags: ai large language model intelligent voice application host, China ai large language model intelligent voice application host manufacturers, suppliers, factory
