1 · A desktop application
The brain and the personality. It runs a local model, does all the thinking, and drives the companion — its expressions, its movement, and its voice.
Pre-launch · in active development
A warm little companion for your desk. Private by design.
A desktop app that gives an off-the-shelf StackChan kit a face, a voice, and a sense of timing. Everything runs locally — there's no cloud to send your data to.
Bring your own StackChan kit — we make the software, not the hardware.
How it works
Software first. Hardware optional.
Lumini v1 is two parts: a desktop app that does the thinking, and an off-the-shelf kit that gives it a face and a voice.
2 · An off-the-shelf StackChan kit
The face and the gestures — a screen-face on a body that can turn its head. You buy it from existing open-hardware sellers and assemble it yourself. We don't make or ship hardware.
What is StackChan?
An open-source, off-the-shelf desktop robot kit — a small screen "face" on a body that can turn its head. It's widely available from open-hardware sellers and assembles in an afternoon. Lumini connects to a compatible kit you already own or buy yourself, and a linked DIY build guide lowers the barrier for newcomers.
What it does
A companion that actually reacts to you.
Lumini sees your expressions, hears your tone, and remembers your context — a gentle hint when you're stuck, quiet company when you're focused.
A face with feeling — driven from your desktop
The desktop app reads your rough emotional state through your computer's camera and microphone, then responds in character — a gentle look, a word, a turn of the head. The robot is the expressive body; your PC does the thinking. It stays put and turns its head — a companion, not a rover.
Empathy for your screen — with permission you control
With a permission you grant and can revoke at any time, Lumini notices what you're doing — racing a deadline, studying, creating — and reacts: a hint when you stall, a small celebration when you break through. The capture happens on your machine and is never uploaded.
It learns your context — locally
A lightweight learning loop runs quietly in the background, building a private picture of what you care about and how you like to work. Everything it remembers is stored locally, inspectable, and deletable by you. We don't claim it becomes an expert in your field — it learns your context, and we can show exactly what it remembers.
Yours to shape
Local-model-first — and you can swap in your own open-source model. Lumini's light color, expressions, voice timbre, and reaction rules are all open to personalize. Your little light grows, bit by bit, into a companion that's unmistakably yours.
Privacy
Privacy isn't a promise. It's the architecture.
In v1, your emotions, your screen, and your conversations never leave your machine — because there's no cloud to send them to. Lumini's light shines on you, never exposes you.
Everything runs locally
All the thinking happens on your own computer. In v1 there is no cloud to send your data to — privacy is a property of the design, not a pledge.
Screen-awareness needs your permission
Watching your screen requires an explicit permission you grant — and can revoke at any time. When on, the capture happens entirely on your machine.
Inspectable, deletable memory
What Lumini remembers is stored locally. You can inspect exactly what it keeps, and delete any of it whenever you choose.
Bring your own open model
Lumini is local-model-first — swap in your own open-source model. Nothing proprietary you're locked into.
Why trust Lumini
Honest signals, not borrowed ones.
Lumini is a pre-launch, one-person studio. Instead of badges and metrics we can't yet substantiate, here's what's true today.
Private by architecture
Everything runs locally; in v1 there's no cloud to send data to. It's a property of the design, not a promise.
Open & compatible
Runs on off-the-shelf, open-hardware StackChan kits; bring your own open-source model. Nothing proprietary you're locked into.
Inspectable local memory
What Lumini remembers is stored locally, viewable, and deletable by you — no hidden profile in the cloud.
Bring your own model
Local-model-first and model-agnostic. Use an open-source model you choose — and personalize how Lumini looks and reacts.
Free to start
Download the app free; no purchase to evaluate it. Connect a kit you own, and a linked DIY build guide helps if you're new.
Built transparently by one founder
Honest about scope. The "deliberately not in v1" roadmap is published, not hidden — what's planned and why is right here.
Deliberately not in v1
What v1 earns the right to build.
These aren't abandoned — they're the funded-team roadmap, earned by proving people love and will pay for a local, private companion first.
No custom mobile robot (yet)
A robot that drives itself around your desk means a ground-up hardware company — chassis, motors, certification, returns — out of reach for one person. v1 uses a stationary off-the-shelf kit instead.
No cloud sync / roaming (yet)
It would contradict the local-first privacy promise. It's the right paid feature to add only after end-to-end encryption is built and audited.
No marketplace or health services (yet)
Regulated activity and third-party vetting are a compliance burden a solo founder can't carry responsibly yet. Revisit with a team and proper counsel.
The plan
Free to start. Pro stays local.
Here's the planned business model. Nothing is for sale yet — the only action on this site is joining the waitlist.
Lumini app — free to start
Download the desktop app and connect any compatible StackChan kit (bought from open-hardware sellers, not from us). No inventory, no shipping. A linked DIY build guide helps newcomers.
Lumini Pro — planned, local-only
Richer learning and longer memory, premium expression/voice/asset packs, advanced reaction rules, and priority updates — all still running on your own machine. You'd pay for a companion that keeps getting better at being yours, not for access to a cloud.
FAQ
Honest answers
The questions we'd ask too, answered straight.
Is my data really private?
Yes. In v1, everything runs locally on your own computer — your expressions, your screen, and your conversations never leave your machine, because there's no cloud to send them to. It's a property of the architecture, not just a promise. What Lumini remembers is stored locally, and you can inspect or delete it.
Do I need to buy hardware?
To get the full companion (a physical face that turns its head), you connect an off-the-shelf StackChan kit — open hardware you buy from existing sellers and assemble yourself. We don't make or ship hardware. The desktop app is the core, and it's free to start; a linked DIY build guide helps if you're new to it.
What is StackChan?
An open-source desktop robot kit — a small screen "face" on a body that can turn its head. It's widely available and assembles in an afternoon. Lumini is the software that brings a compatible kit to life.
Will Lumini become an expert in my field?
We won't claim that. What we do claim: it learns your context — your topics, your rhythm, how you like to be helped — and gets more useful to you over time. And we can show exactly what it remembers, because it's all stored locally and inspectable.
Can I use my own AI model?
Yes. Lumini is local-model-first and you can swap in your own open-source model. Its light color, expressions, voice, and reaction rules are open to personalize, too.
Is it watching my screen all the time?
Only with a permission you explicitly grant — and can revoke at any time. When on, the screen-awareness happens entirely on your machine and is never uploaded.
When can I use it?
Lumini is pre-launch and in active development. Join the waitlist to get the early build, the DIY StackChan guide, and honest progress notes.
How much will it cost?
The desktop app will be free to start. A planned Lumini Pro subscription will unlock a richer local experience (more memory, premium packs, advanced rules) — still running entirely on your own machine. Nothing is for sale yet.
Pre-launch
Join the waitlist
Lumini is in active development. Leave your email to get the early build, the DIY StackChan guide, and honest progress notes — no spam, no sharing.