Houjun Co., Ltd. · AI research and development · Japan
Private AI,
on your own hardware.
We make AI models run on your own Macs, workstations and on-premise servers — and turn them into software people can actually use. Our open-source tools are the proof: download them and try.
Written first · English, Chinese or Japanese · Fixed scope, fixed price
Nothing leaves the room.
Our public work
Each of these runs locally on an Apple Silicon Mac. Nothing is uploaded.
AppOpen source · AGPL-3.0
VoxStage
Offline multi-voice audiobook and dubbing workstation. Every generated sentence is transcribed back and checked against the script.
Visit VoxStage →
AppOpen source · MIT
Image MLX Lab
Local image generation and masked AI editing. Only the selected area changes; every pixel outside it stays untouched.
Visit Image MLX Lab → Research2 projects completedHoujun Lab
Small experiments, measured honestly: what a small language model learns from new papers and what it forgets, and how a detector and tracker really perform on a CPU.
Visit the Lab →What we do for clients
Four ways to start. Scope, acceptance criteria and price are agreed in writing before any work begins.
Feasibility review
1–2 weeks. We test candidate open-weight models on your hardware and your sample data, and tell you plainly whether to go ahead.
Proof of concept
3–6 weeks. A working prototype on your data, running where your data lives.
Training and adaptation
Fine-tuning open-weight language models on your own documents, with the evaluation designed first.
Licensing and integration
Commercial licences for VoxStage, and help building our tools into your product or workflow.
A good fit when
- Your data cannot leave your machines — medical, legal, unpublished or customer data.
- Cloud API costs grow with every request, and you want a fixed-cost alternative.
- You need AI that works offline or inside a closed network.
If a cloud service already solves your problem well and cheaply, we will tell you.
Since 1998
Our lead engineer has built intelligent systems since 1998: lung-sound recognition with a university hospital, large-scale real-time video, robot vision, and now generative AI running locally.
About us →
We ship our own products, too
Apps on the App Store, and a coffee roastery in Tokyo.
Mind Craft Fish
A learning ocean for children: short games grow an aquarium and a beach of their own.
One Line Shift
A one-line drawing puzzle. Cover every path in one continuous line.
Cherry Tempo
Reminders and shared task lists, clearly and calmly.
Cherry Lab
Specialty coffee, roasted to order.
Tell us what you want AI to do
Email us with the subject “Project enquiry”. We reply within two business days with a written first assessment.