A company AI that feels like talking to a real person.
From reception and tourist information to web support, events and education. Look, voice and personality are designed for your brand, and it answers 24/7 in Japanese, English, Chinese, Korean, Spanish, German and 100+ other languages using only the official information you register. The new 100% local version runs on a single PC with no internet, and is built in under an hour.




See real conversations in action.
The demo we presented at AI BEYOND 2026 (about 90 seconds). Live-action, anime and 3D characters answer in Japanese, English and Spanish using only the information you register.
MetAI Tomo comes in four editions.
The little flame at the bottom right of this site is a live example of the Web (Lite) edition. Choose the one that fits your use and budget.
MetAI Tomo Local
Everything runs inside one PC — 100% local. No internet needed: a live-action, anime or 3D look, your own voice and lip movement, generated in real time for every answer.
- No internet · data never leaves your site
- Real-time voice and lip-sync for every answer
- No answer limits (within the hardware)
- Recommended for hospitals, banks and local governments that handle confidential data, and for companies that want an AI face at reception, showrooms and trade shows
MetAI Tomo Web (Lite)
A lightweight edition you simply add to your website. It runs on cloud AI, so no dedicated hardware is needed — and the Tomo on this site runs with no separate API or server costs.
- Instant spoken answers to common questions, text for the rest
- Voice input, multiple languages, chat log and admin panel
- Need lip-sync or unlimited answers? Step up to Hybrid or Local
- Recommended for companies, local governments and schools that want AI at reception, inquiries and events on a modest budget
MetAI Tomo Hybrid
A local agent running on your own PC, connected to your website. Questions from the site are answered live, every time, in your own voice.
- The convenience of Web with the voice of Local
- Confidential data stays on your own PC
- Works while the PC is on (text-only when it’s off)
- Recommended for small businesses, clinics, professional offices and schools that want to answer website inquiries in a real voice while keeping their data in-house
MetAI Tomo Cloud (100% cloud)
Everything runs on cloud GPU servers. No PC is needed on your side — even real-voice and video conversations run fully online. MetAI builds and runs it for you.
- No hardware on site · the whole service online
- Ongoing cloud server costs
- Recommended for companies with the budget that want everything online
* The Tomo on this site is a Web-edition demo customized for MetAI, and it differs from the setups we build for clients. We can create something similar for your website, and we’ll suggest the edition that best fits your budget and needs.
Four types to choose from: live-action, anime, 3D, full-body
All of them are fully local and run without internet, on the same RAG base with multilingual, real-time conversation.
Shown at the EXPOLive-action
Talks with the person’s own look, voice and manner. For tourist guidance and concierge use.
Shown at the EXPOAnime
Anime-style characters (Ana, Pan). Speak with voice, blinking and lip movement.
Shown at the EXPO3D character
A full-body 3D-style character (Taka). Swap by simply calling a name.
Shown at the EXPO3D + agent
A flame-shaped 3D character (Tomo). Controls apps, volume and more on the PC by voice.
* All four types debuted as prototypes at NIIGATA AI EXPO 2026 (Oct 28–29). From December 2026 they will be field-tested at NINNO3, next to Niigata Station.
![[object Object]](../../assets/vh2026/demo_3idiomas.webp)
Japanese, English, Chinese, Korean, Spanish, German and 100+ more languages.
It detects the language it is spoken to in and answers in the same one, so international visitors are served in their own language. Backgrounds and character voices can be changed with one tap.
- Answers from official information only (RAG); never guesses and points to a contact
- Voice: reproduce the person’s own, a character voice or one you prefer
- Conversation logs and visitor feedback organized automatically for improvement
- Optional: embed on a tourism site or use through a QR code
From API-based to 100% local.
Speech recognition, answer generation, voice synthesis and lip movement all run on your own PC. With no waiting for the cloud, both creation time and response time drop sharply.
Avatar creation
About 1 month → 1 day → under 1 hour (typical → our earlier method → now). Live-action starts from a 2–3 minute recording.
Response time
2–10 s → 1–6 s → under 1 s (from question to start of reply). No network wait, so the conversation feels natural.
Cost and data
One-time purchase, no API fees. Conversation data is processed only inside the PC and never sent to the cloud.
| Item | Until now (API version) | From now on (100% local) |
|---|---|---|
| Environment | Internet required | No internet, a single PC |
| Response time | 1–6 seconds | Under 1 second |
| Avatar creation | 1 day | Under 1 hour |
| Cost | Ongoing API fees | One-time purchase, no API fees |
| Conversation data | Sent to the cloud | Processed only inside the PC |
* Figures come from our own tests on an HP workstation and are not independently verified. Typical figures for conventional services: about 1 month to create and 2–10 s to respond. The product is under development; a prototype debuted at NIIGATA AI EXPO 2026.
Data stays inside
Conversations and company documents never leave the PC.
Works offline
Even in the mountains, at venues or where signal is weak.
No API fees
Costs do not grow with usage.
Fast replies
No network wait: under 1 second.
More useful, for everyone who uses it.
The 100% local version now adds access control, each company's own knowledge, a usage dashboard, conversation logs and answers with images.
Four types of access
Guests, users, companies and administrators each get the right permissions. Companies can sign in only with an access code issued by the administrator.
Each company's own knowledge
Add a URL or Excel, Word, PowerPoint, PDF and CSV files, and it answers with that company's information. The data stays on the company's own PC.
Usage dashboard
Conversations, time of use, usage by avatar and by user, questions by hour, and the questions it could not answer, at a glance.
Conversation log
Every conversation, searchable by user, avatar or keyword.
Answers with images
When there are several answers, each one appears in the chat as a card with an image, date, place and link.
More than 10,000 pages learned
Even with more than 10,000 pages learned, the first answer arrives in about 1 second. There is no limit on the number of pages.
8 ways to use it
The videos and images shown are the previous API-based models. The 100% local versions (live-action, anime, 3D) are available for demos on request.
![[object Object]](../../assets/vh2026/demo_3idiomas.webp)
Tourist information
Built on a tourism-specific RAG for Niigata. It guides visitors in several languages using only registered official information, even outside staffed hours or where signal is weak.
- Eases staff shortages and multilingual needs at tourist offices
- Prevents wrong guidance and protects trust
- 24/7 guidance by website embed or QR code (optional)

Store & reception
Covers the work of two receptionists (¥6M+/year). No waiting, multilingual for international visitors.
- Several visitors at once
- ¥4.8–16M saved per year
- Excel export of inquiries

Website concierge
Face-to-face service by video call — recommendations that link straight to product pages.
- 24/7 multilingual
- Links to product pages

Events & trade shows
Visitors just ask “Where is X’s booth?” on their phone — and get briefed on the company on the way.
- A day’s visit in half a day
- Twice the leads

Elderly care
Someone to talk to 24/7 — and if the camera detects a problem, it alerts hospitals or emergency services.
- Medication reminders & check-ins
- Updates for family

Education
24/7 study help and guidance — students who can’t attend or study abroad learn from home in their own language.
- Question analytics improve lessons
- Supports students at home
![[object Object]](../../assets/vh2026/char_taka.webp)
Characters & IP promotion
Let fans talk to a local mascot or IP character. It only says what the rights holder has approved, so it is safe to use for promotion.
- Choose live-action, anime, 3D or full-body
- Dedicated character voice and backgrounds
![[object Object]](../../assets/vh2026/char_tomo.webp)
AI agents
Beyond conversation: an AI agent you can hand your PC work to. Just ask by voice or chat and it gets everyday office tasks done for you. We are rolling out support for tasks like these (a prototype already controls apps and volume by voice).
- Scheduling and meeting booking (calendar integration)
- Sorting email and drafting replies
- Writing, scheduling and replying on social media
- Meeting minutes and task follow-up
- Quotes, invoices, expense reports and other documents
- Data entry into Excel and internal systems
- First-line replies to inquiries, handed over to staff
- Reports on sales, stock, web traffic and more
- Research from the web and internal documents
- Guiding staff through procedures in your manuals
- Your company data stays on your own PC

“Sato from HP” — a showroom specialist, turned into AI.
Built specifically for HP, recreating a showroom specialist’s look and manner. It recommends the right model based on each visitor’s work, PC use and budget — live at HP Japan’s HQ showroom.
- Specialist recreated as AI
- Answers product questions on the spot
- Recommends within budget
- 01
15 s · Listen
Silent, as if listening
- 02
90 s · Talk
Any topic
- 03
15 s · Hold
Stay still
A look-alike avatar from a short recording.
It learns your way of speaking, tone and gestures, with lip-sync. After that, just type text to create videos.
Selected for a Niigata Prefecture subsidy. First shown at NIIGATA AI EXPO, Oct 28–29.
Project: “Development and pilot of an on-premises MetAI Tomo with a Niigata tourism-specific RAG”. Selected under the FY2026 AI Utilization Promotion Lab subsidy (NIIGATA AI Digital Playground, hosted by Niigata Prefecture).
NIIGATA AI EXPO 2026
Wed Oct 28 and Thu Oct 29, 2026 · Bandaijima Multipurpose Plaza “Okama” (Niigata City) · MetAI Booth 54. First public showing of the prototype: visitors tried the conversation, character/voice/background switching, multilingual chat and the conversation log.
Pilot: NINNO3
From December 2026 to February 2027 it will be placed near the reception of the NINNO3 meeting rooms (Plaaka 3, next to Niigata Station South Exit) and validated with visitor surveys. Expected visitors: local governments, IT companies, creators, out-of-prefecture companies and investors.
After the pilot: across Niigata
After the pilot, deployment to tourist facilities and companies in Niigata will proceed through ASK Corporation as the point of contact.
- 2026.10Oct 28–29: prototype shown for the first time at NIIGATA AI EXPO 2026 (Booth 54)
- 2026.12Review of the finished version with ASK Corporation and HP Japan
- 2026.12–Dec–Feb 2027: pilot at the NINNO3 meeting rooms — operated by Kiyama Sangyo Co., Ltd., managed by Kento Housou Inc.
- 2027.02Pilot results reported; deployment begins (contact: ASK Corporation)
MetAI LLC
Develops the virtual human, builds the RAG, multilingual and voice support, remote maintenance
ASK Corporation
Sales and support: finding customers in the prefecture, sales contact and first-line support
HP Japan
Hardware: AI workstations and hardware maintenance
Kiyama Sangyo Co., Ltd. & Kento Housou Inc.
Pilot cooperation: placement at NINNO3 and visitor surveys
See how it went in our EXPO report. For press or demo requests, use the contact form.
Compared with a typical chatbot
Japan’s AI chatbot market is heading for ¥150 billion by 2030 (22% CAGR). MetAI stands out for human-likeness and speed of deployment.
| Feature | Typical chatbot | MetAI virtual human (API version) |
|---|---|---|
| Voice | Robotic voice | Natural human voice |
| Appearance | Mechanical | Real expressions & gestures |
| Knowledge | Data-leak risk | Only your company’s data |
| Channels | Text only | Voice and text |
| Time to launch | 6+ months of development | 1–2 months (varies with requirements) |
| Avatar creation | About 1 month | 1 day |
Mode 1 · Closed
No external AI — answers only from your data, and nothing leaves the company. The same accurate information as your staff, faster.
Mode 2 · Hybrid
Your data plus real-time web search — weather, news, latest prices — with the search scope under your control.
How much could you save?
Adjust staff count and annual cost per person (estimate at up to 80% savings).
* Rough estimate; actual savings depend on your operations and setup.

The demo lives right here — bottom right.
Tomo, the little flame, is the Web edition of MetAI Tomo. Ask about our services, pricing or what fits your industry — by voice or text, any time.
* Demos of the Local and Hybrid editions — live-action, anime or 3D, with a real voice and lip-sync — are shown during the free consultation.
