The AI Travel Companion
That Tells the Best Travel Stories
Tome AutoRoam independently developed the Tome-ubind vertical large language model for tourism, with parameter scales from 7.5B to 20B, giving every journey an intelligent narrator who understands history, culture, and you.
Tome AutoRoam AI Reinvents the Tourism Experience
Empowering consumers
with immersive AI experiences,
while multi-agent systems drive business growth.
Independently Controlled Technology
Fully self-owned LLM technology, featuring a unique pipeline of pre-training + vector training + RLHF (Reinforcement Learning from Human Feedback).
Deep Operational Experience
Independently developing and deploying smart tour-guide devices since 2017, now operating in 300+ locations domestically and 11 countries abroad — an industry leader.
Proven in Real-World Scenarios
Successfully deployed at numerous 5A-rated scenic spots and museums nationwide, validated through large-scale pilot operations.
Clear Business Model
Starting from sticky tool usage, rapidly scaling online traffic, and driving sustained revenue growth through diverse value-added services.
Seven Years of Dedication, Advancing Toward AGI
From smart tour-guide hardware to a self-developed tourism LLM, Tome AutoRoam continues to invest in R&D and industry-academia collaboration.
First-Generation Smart Tour Guide Devices
Independently developed 1st–3rd generation smart tour-guide devices (Tome-A1/A2/A3), deployed across scenic spots to build user habits.
Launched Industry-Academia Collaboration
Partnered with Sichuan University to establish the "China Tourism & Big Data Research Center", jointly exploring a new smart tourism ecosystem.
Launched Computer Vision R&D
Co-founded an AI research team with the School of Computer Science at Chongqing University of Technology, initiating computer vision research.
Launched NLP R&D
Initiated natural language processing research, exploring LLMs and Transformer-based pre-trained models.
Selected for Tencent AI Accelerator
Tome Vlog was selected among the top 8 of the Tencent AI Accelerator, co-branding "Tencent Tourism × Tome Vlog" with Tencent.
First NLP Model Released
Officially released the company's first NLP model, Tome-ubind v1.0.
Tome-ubind v5.0
Released the self-developed 7.5-billion-parameter model Tome-ubind v5.0, with significantly improved performance.
Selected for International Accelerator Programs
Selected for the AWS China and NVIDIA Inception startup accelerator programs, officially advancing toward AGI.
Tome-ubind — Vertical LLM for Culture & Tourism
Tome-ubind is a pre-trained LLM built with cutting-edge NLP technology, trained with Reinforcement Learning from Human Feedback (RLHF) for Chinese-English language preferences, with parameter scales ranging from 7.5B to 20B. Powered by contextual temporal attention and continuous sentiment recognition, the model supports multi-turn conversation memory, self-correction, and continuous self-learning.
The model has been quantized for deployment on standard GPU infrastructure, offering low inference costs and high efficiency. It also supports private deployment and targeted data training, keeping data on-premises for full security and compliance.
| Evaluation Dimension | Tome AutoRoam LLM | General LLM A | General LLM B |
|---|---|---|---|
| History – Timeline Completeness | 92 | 65 | 68 |
| Humanities – Sentiment Analysis | 89 | 63 | 76 |
| Legends – Version Tracing | 95 | 72 | 68 |
| Science – Cross-disciplinary | 88 | 75 | 80 |
Scores are out of 100. Data based on Tome AutoRoam's internal evaluation system and quantitative benchmarking against mainstream general-purpose LLMs.
From Heavy Hardware to a Lightweight AI Travel Companion
Moving beyond one-way broadcast tour-guide devices, Tome AutoRoam redefines smart narration with digital-human AI agents.
Traditional Hardware Tour Guide
- Heavy asset, one-way broadcast, no interaction
- Heavy offline operations (rental staffing, hardware maintenance)
- High operating costs; revenue sharing required with partners
Tome AutoRoam AI Companion
- Lightweight, digital-human narration, chat-based interaction
- Fast online promotion with broad coverage
- Low operating costs, no revenue sharing
An Immersive Travel Experience Like a Personal Guide in Your Pocket
Serving tourists, scenic spots, and agencies with a full capability matrix covering narration, itineraries, services, and private-domain operations.
AI In-Depth Narration
- Seamless GPS-triggered narration at points of interest
- Real-time AI Q&A (RAG) to answer visitor questions anytime
- Emotional voice synthesis with multiple characters, dialects, and languages
- Multimodal visual recognition — identify landmarks and artifacts by photo
Personalized Itineraries
- Dynamic routing based on preferences, weather, and crowd levels
- LBS-based gamified exploration
- AR historical reconstruction of ancient architecture and heritage sites
- One-click cinematic AI travel vlog generation and sharing
Omni-Channel Service Gateway
- Voice-activated services in one sentence — directions, ticketing, ordering
- Direct API integration with merchant systems for automated fulfillment
- Branded, customized solutions for scenic spots and merchants
Private Domain Operations
- Customizable digital-human guide (avatar, voice, greeting)
- Exclusive curated routes with 24/7 companion service
- Lead conversion dashboard with real-time consumer intent insights
AR Time Envoy · Multimodal AI Agent
Built on a private large language model, the "Time Envoy" integrates visual RAG technology to match visitors' gaze focus with historical and cultural databases in real time — pioneering a closed loop of "AR + deep LLM + visual recognition interaction + semantic parsing + emotion computing".
AR Immersive Cultural Exploration
Wearing AR glasses, visitors step into historic sites as the AI agent tracks gaze focus in real time and overlays 3D reconstructions — bringing millennia-old craftsmanship to life.
Emotion-Aware Narration
Using eye-tracking and micro-expression recognition, the agent dynamically adapts its narration style — playful storytelling for children, academic discussion for researchers.
Spatial Navigation Cognitive Enhancement
Recognizes signage and exhibit text, translates it instantly, and overlays cultural background — turning "understanding" into "remembering".
Smart Tourism Results, Backed by Data
Validated through large-scale pilot operations at numerous 5A-rated scenic spots and museums.
Scenic spots and museums with deployed smart tour-guide pilots
Increase in average visitor dwell time at a 5A-rated classical garden pilot project
Improvement in visitors' cultural knowledge retention at the same pilot project
Narration corpus covers 1,000+ domestic scenic spots and 30+ countries abroad; data based on Tome AutoRoam's internal pilot project statistics.