Introduction
On September 1, 2026, Dyson announced the CameraJet, a $499 electric toothbrush with a camera in the head and a machine vision model deciding, several times a second, where to squirt mouthrinse. The pitch is that it replaces floss: instead of you finding the gaps between your teeth, the toothbrush finds them and jets them clean. It is on sale now in Ceramic Ultra Blue and Ceramic Pink.
Whatever you make of the price, the CameraJet is an unusually clear example of computer vision shipped as edge AI in a mass-market consumer object — a real-time perception loop running inside something you hold in your mouth for two minutes.
How the targeting loop works
A 100,000-pixel macro camera sits under the bristles with its own LED illumination. The brush head moves close to 1,000 times per second, which would smear every frame, so light and shutter strobe slightly out of sequence with it — at 257 Hz, according to outlets briefed by Dyson; the figure does not appear in the press release — letting the camera catch usable frames between bristle sweeps.
Those frames feed Gap Optical Targeting, Dyson's machine learning model trained on 470,000 dental images gathered during development. It runs inference over 28 frames per second, identifies an interdental gap, and fires a conical jet of mouthrinse into it within about 100 milliseconds. Dyson says that analysis happens on the brush itself rather than on a paired phone or a server — which the timing budget makes plausible, since a 100 ms round trip to the cloud would leave nothing for the actual detection. The 12.5 ml tank uses a diaphragm pump and pressure chamber so the jet keeps priming at any angle you hold the handle. Dyson quotes six years of development, 661 engineers and 16 million lines of code, and claims, on its own testing, 69% more plaque removal in hard-to-reach areas than leading premium electric brushes — without naming the comparison models or publishing the protocol.
The consumables are part of the system
Two engineering constraints turn into recurring purchases. Foam blinds the camera, so Dyson made a non-foaming, SLS-free toothpaste; the jet needs a low-foaming mouthrinse formulated for it. The dock refills the tank in three seconds. Brush heads carry RFID tags for tracking. You can also stream the camera feed to the MyDyson app over Wi-Fi and use the brush as an ad-hoc dental endoscope, with coverage maps and feedback on the spots you keep missing.
Dyson's privacy claim is narrow but specific: the camera runs only during live viewing or automatic jetting, and images are neither recorded on the brush nor sent to the cloud. Nothing here is independently audited — it is a vendor statement about a camera pointed inside your body, and worth reading as one.
Conclusion
The interesting part is not the toothbrush, it is the engineering budget a consumer appliance can now justify for a perception model: a bespoke strobe frequency to beat motion blur, a proprietary toothpaste so the lens can see, and a 470,000-image training set — all to solve flossing. Whether $499 plus consumables is the right price for that is a separate question from whether the technique works, and independent reviews of the plaque claims have yet to land.
Sources
- Dyson CameraJet press release (PR Newswire)
- Introducing CameraJet — Dyson
- Dyson's CameraJet is a $500 three-in-one toothbrush that scans your maw — Engadget
- Dyson puts a camera and a jet in a toothbrush — techAU (source for the 257 Hz strobe and the on-device processing claim)
- Dyson Put a Camera in a $499 Toothbrush — TechRepublic