One operating system for everything a child holds. You build the body. PlayOS brings the soul.
Two experience families. Six tiers. One module.
Per-unit pricing at design-partner volumes, in USD (₹ quotes for India programs on the call). The exact number lands on your spec call: it moves with options and order size, not with our mood.
Companion module
Talk & Respond
Wake word, questions and answers, stories and songs. Voice runs in the cloud.
Real Conversation
Continuous dialogue with memory and 10+ languages; wake-word on the device.
Sees & Understands
Everything above, plus vision: show-and-tell, object play, real presence.
Baby-care module
Monitors & Alerts
HD IR camera, cry detection that knows hungry from sleepy, room-temperature alerts.
Soothes & Responds
Adds AI soothing: lullabies and two-way audio that answer before parents are even up.
Sleeps & Learns
Adds sleep analytics, early-wake detection and motion control for auto-bounce cribs.
The Magic Box: PlayOS, in a box.
A palm-sized module (custom PCB, MCU, mic array, speaker) running PlayOS. Drop it into any shell and the shell comes alive.
The same module powers Lumi, Lori and every partner product. When PlayOS gets smarter, so does your shelf.
Parents see everything. That's why they keep it.
Every PlayOS product ships with a parent app you put your own brand on. It is how a parent knows the toy is helping, not just talking, and it is the reason they renew, month after month.
Summary & notifications
An actionable summary of the day, and a nudge the moment something matters.
Conversation log
Every conversation, in full. Parents read exactly what their child talked about, anytime.
Filter topics
Parents choose what the toy talks about, and what it never will.
Reinforce cultural values
Ask the toy to weave a family's language, culture and values into play.
Parenting philosophy, in one prompt
Align the companion to how a family parents, in a single line of text.
Parent it your way, in one line.
A parent types how they want their child guided, and the toy follows. Ask it to make patience the lesson, with the child as the hero of the story. Their values, their culture, their words.
It turns curiosity into a next step.
A morning spent talking about birds becomes a suggestion to visit the bird sanctuary this weekend. The app surfaces the real-world moment the conversation opened up.
The parent app ships with every PlayOS product, branded as yours.
A child speaks. The toy answers. Here's the path in between.
Wake, or press
The device listens only on a wake word or a button. The rest of the time the mics are off. Better for privacy, better for battery.
Capture
The mic array picks up the voice, and on-device voice-activity detection trims the silence before anything travels.
Edge or cloud
Light turns can run on the module itself. Richer turns route to the PlayOS cloud. The same firmware decides, per device, per tier.
Voice to voice
The Voice SLM goes from voice straight to voice, with no text round-trip in the middle. That is what keeps a reply quick enough that a three-year-old doesn't lose interest.
Safety on every turn
On-device filters and cloud guardrails screen what goes in and what comes out, grade the answer to the child's age, and keep it off the open internet.
Answer
The toy speaks back in the child's language, remembering what was said a minute, or a week, ago.
Exact latency targets are spec'd to your tier and hardware on the call.
The full stack. You drop in one module.
Six things stand between a child and a bad answer.
The smaller the brain, the smaller the blast radius. A sub-1B model has far less surface to go wrong than a frontier one, and it never talks to the open web.
Pinned to a region. Deletable in a tap. Never sold.
No ML team. No audio pipeline to write. No compliance to invent.
A certified firmware image
Audio pipeline, safety, OTA and device management already inside.
An SDK
Add your own language packs, curriculum, voices and characters.
A content pipeline
New stories and fixes, pushed over the air.
Parent-app endpoints
Config, progress, controls and SOS, branded as yours.
Our engineers
Hands-on for your first integration.
From call to shelf in 4–6 weeks.
On our call we pick the tier, the hardware and the per-unit module price for your product.
We deliver the module (PCB, MCU, mic array, speaker) sized to your tier.
Certified firmware drops onto your line; the SDK connects cloud, app and content.
Your shell, your IP, your shelf, on sale in 4–6 weeks.
The Voice SLM: the voice brain, as an API.
A sub-1B-parameter voice-to-voice model (no text in the middle), built for the latency and edge constraints of real toys, not datacenter assumptions. Partners get access from Q4 2026.
Training data comes only from conversations families separately opt into: anonymised, withdrawable with one tap, never sold. Every consented conversation makes the model smarter.
Voice SLM
Kids talk to our toys
Real conversations, every day
Data feeds the Voice SLM
Consented, kid-safe signal
The model gets smarter
Better voice, better answers
Every product gets better
Which makes kids talk more