On-device attention verification

Know when someone is actually watching.

Optivise uses the camera already on the device to confirm real human attention. The instant focus drifts, your content knows, whether it is an ad, a training module, a lesson, or the road ahead.

Try the live demo See how it works
GAZE LOCKED
ATTENTION 98%

A screen counts a view the moment it plays. It has no idea whether a person ever looked.

~2 of 3
video plays happen with eyes somewhere else from multitasking, a second screen, or an empty room.
0
ordinary screens that can tell a watching face from one that walked away.
1.7s
average time attention holds on a screen before it drifts elsewhere.

Attention, verified frame by frame.

The whole pipeline runs in the browser or app on the viewer's own device. No special hardware, just the camera they already have.

01

Detect

A lightweight model locates the face and eyes in the camera feed at 30 frames per second, entirely on device.

02

Verify gaze

Optivise estimates head pose and eye direction to decide, in real time, whether the viewer is genuinely looking at the screen.

03

React in real time

The instant attention drops, Optivise responds by pausing playback, flagging the moment, or alerting on the spot, or other customizable action. When focus returns, it picks up exactly where it left off.

04

Report

Every session produces a true attention score: seconds watched, look-aways, and completion that reflects real human focus.

See it work on your own face.

Enable your camera and watch a sample ad. Look away and it stops. Look back and it continues. Nothing leaves your device.

Launch the live demo
Use cases

One attention engine. Wherever focus matters.

The same on-device gaze model adapts to whatever is on the screen. Fan through the surfaces it works on, advertising is simply where it started.

01

Advertising

Bill for attention, not impressions. Pay only for the seconds a real person actually watched.

$84B
spent each year on impressions that may never reach a human eye.
100%
of billed seconds are watched seconds with Optivise, verified on the device.
2.7x
recall lift versus ordinary, unverified views.
02

Workplace training

Compliance and onboarding videos log a completion only when an employee actually watched.

1 in 3
mandatory training videos finish with no one watching the screen.
100%
of logged completions reflect genuine, eyes-on watch time with Optivise.
0
extra hardware. It runs on the webcam already there.
03

Schools & e-learning

See who is really following a remote lesson and taking tests with integrity. A true picture of focus, without recording a child.

2.4x
clearer view of engagement than attendance or quiz scores alone.
85%
reduced chance of cheating with Optivise.
0
faces stored. Only an anonymous focus signal is ever produced.
04

Long-haul delivery

Eyes on the road, all night, the gaze model flags drift and drowsiness before it matters.

<0.5s
to flag eyes leaving the road and warn a fatiguing driver.
24/7
on-device monitoring that runs in the car with no connection needed.
0
footage uploaded. Alerts driver locally and nothing is recorded.
05

Online meetings

Engagement analytics for video calls, see when a room is genuinely tuned in, not just logged on.

62%
of the average call is spent looking away, on another tab, or off-camera.
live
per-participant attention so hosts can tell the moment a room drifts.
0
video stored, only an aggregate engagement score ever leaves the device.
06

Contract verification

Prove a signer actually read the document. Gaze tracking logs which sections were viewed before the signature is accepted.

91%
of people admit to signing contracts they haven't fully read.
section-level
gaze coverage map shows exactly which clauses received attention.
0
document content ever leaves the device. Only the attention log is stored.

Don't see your use case? The attention engine is fully adaptable: custom triggers, thresholds, reporting formats, and integration points. Reach out to the team to talk through what you need.

Two modes

Pick the right level of precision for the job.

Both modes run entirely on the viewer's device. The difference is how precisely Optivise knows where on the screen the viewer is looking.

01 / Quick start No setup

Attention verified, no configuration.

Works the instant the camera turns on. Fixed gaze-angle thresholds calibrated to a typical viewer determine whether someone is looking at the screen. No personal setup required.

  • Setup timeNone
  • AccuracyGeneral — ±8–12% error
  • Gaze outputAttention on screen / off screen
  • Best forAds, meetings, general engagement
Try quick start
02 / Calibrated ~30 s setup

Exact gaze position on screen.

A two-phase personal calibration: a static nine-point grid followed by a smooth-pursuit sweep. Builds a gaze model tuned to this viewer's eyes and posture. The result is a live cursor tracking exactly where they are looking.

  • Setup time~30 seconds, once per session
  • AccuracyPersonal model — ±3–5% error
  • Gaze outputPrecise X / Y position on screen
  • Best forTraining, contracts, proctoring
Try calibrated

One layer, every screen with a camera.

Web and players

An SDK that drops into any HTML5 video or web app whether its ads, courses, or lessons with only a few lines of code.

Mobile and tablet

Front-facing cameras make in-app video, training, and classroom apps truly attention-aware.

Embedded and in-car

Runs on the cameras built into kiosks, vehicles, and dedicated hardware, fully offline.

Privacy is the product.

Eye tracking only earns trust if no one is being recorded. Optivise is built that way from the first frame.

No. All detection runs locally on the device. The video never leaves the browser or app. Optivise only emits a small attention signal, such as "watching" or "looked away," never imagery.
Aggregate attention metrics only: seconds watched, look-away counts, and completion. No faces, no biometric identifiers, no raw frames.
Always. The camera turns on only with explicit consent, and viewers can fall back to a standard playback experience at any time.
Any device with a standard front camera. No depth sensors, no extra peripherals, no calibration ritual.

Make every screen know when it has your attention.

Try the working prototype, then let us bring attention verification to your ads, your training, your classroom, or your fleet.

Try the live demo Talk to the team