peetaboo.com·private beta·Oakville, Ontario

Peekaboo, the way she says it.

Our daughter Ellie calls it peetaboo. PeetaBoo is the home camera system we built so the thousands of small moments like that one, already caught by the cameras in our house, get found, understood, and kept. Not scrubbed through. Not lost in a camera roll.

Join the waitlist Watch the raw footage

Works with the camera you already own. Identity matching runs on hardware in your home.

Ellie, almost two · 2.7 s · the reason for the name

10,690
motion clips scanned from one living-room camera, Feb to Jul 2026
9phases
of identity checks before a clip is called a match
0.90mAP50
for our fine-tuned child detector, up from 0.33 with the stock model
0
false positives across the 134-clip validation batch
Raw footage  ·  unedited

A Tuesday morning at 9:53. Nobody was filming.

This is a real clip from the security camera in our living room. Motion triggered it, PeetaBoo found Ellie in it, described what happened, and I tapped the bookmark so it never expires. The observations beside it are the system's actual output for this clip, word for word.

24 Feb 2026 09:53:08 Reolink, wall-mounted motion-triggered, 11 s bookmarked by Dad

For parents who work, the day happens in the room you're not in.

Whether you're on a call upstairs or across town, most of playtime happens without you. Cameras already record it. Nobody has time to watch it. PeetaBoo watches for you and hands back the eleven seconds worth keeping.

For first-time parents, everything is a first.

The first peekaboo. The first time she climbed the chair. The first "op op op" at a blue cup. You won't have your phone out for any of them. This is how you store them anyway, in a library that stays organized on its own.

How it works

Three engines. One quiet library.

PeetaBoo is not a smarter baby monitor. It is a pipeline that runs on the footage your cameras already produce, built to be right rather than fast.

01  FIND

Is that my child?

A person detector we fine-tuned on ceiling-camera footage finds people and tells children from adults. Face detection and a 512-dimension face embedding then match against 3 to 5 reference photos, and against a gallery of family faces, so a sibling, a parent, or a doll is never mistaken for your child. Nine phases of checks. Runs on hardware in your home.

YOLOv8n, fine-tunedRetinaFaceArcFaceBoT-SORTnegative gallery
02  UNDERSTAND

What happened?

Only clips that match are described by a vision-language model, in plain observations: what she did, who was there, how she seemed, and what a pediatrician might find worth knowing. Low temperature, no clinical labels, no diagnoses. Ever.

vision-language modelnative video inputtemperature 0.2observations, not verdicts
03  KEEP

What's worth keeping?

A daily journal on your phone: the day's key moment, activity breakdown, mood, and clips. Bookmark a clip and it is kept forever. Everything else ages out on a schedule, so you end up with a library instead of a hard drive.

iOS & Androiddaily journalbookmark = keep forevermulti-child
$ peetaboo scan "02-24-2026 09-53 AM.mp4" frames_sampled 37 every 3rd frame, about 10 per second faces_detected 5 quality-gated before embedding identity Ellie matched against her reference photos, checked against family faces verdict MATCH copied to matched/, original left untouched insight 3 segments social_interaction · transitioning · watching keep true bookmarked 2026-04-05, never expires processing 52.7 s on a laptop, no GPU
Privacy

Built by a parent who reads the privacy policy.

This started as software for one family's footage. The rules we set for ourselves are the rules the product ships with.

Be one of the first hundred homes.

We're running PeetaBoo in our own house and a handful of others. If you have a camera and a kid under three, we'd like to hear from you.

Thanks. We'll write back from a real inbox.

Oakville, Ontario  ·  a human replies within a day  ·  hello@peetaboo.com