Everyframe
Give it video. It gives back a result. The parts combine .
Everyframe is our video software, built and hosted in Europe. A few of the things it has already done, on real footage.
Live and public. No sign-up.
One rider stays sharp. Everyone chasing him goes blurry.
Every other face is found, then pixelated. The rider is kept by his track identity, not by face recognition. Nobody drew a box by hand.
369
Faces detected
186 frames
1,410 award videos. Nobody opened an editor.
One design and one list, for a national innovation ceremony. Each video came back finished, in its own region and voice.
235 entries / 6 voice variants / 13 regions
Composer, the app for this job, is coming soon. Pre-beta, no date.
1,410
Finished videos
One cell, one rendered video
Every lift counted. Nobody wearing a sensor.
The body becomes 17 tracked joints. There is no rule per exercise: it finds the repeating motion itself. Given footage that does not repeat, it refuses to count.
Open the showcase →17
Tracked joints
No sensor worn
Vehicles read from the air. The video never leaves the drone.
A Jetson board on the aircraft does the seeing, fully offline. Home gets about 68 bytes per record, never a frame.
Perception, not force. It reads a scene and hands it to a person. A human stays in every decision.
68
Bytes per record
17 W on the board
Old footage, made usable again.
The 1963 Dutch Grand Prix, denoised, deblurred, upscaled and smoothed by the engine's Claritas. Only 1963 is a Formula 1 race; the earlier footage is racing at Zandvoort.
Open the showcase →1963
Dutch Grand Prix
restored, not reshot
Why one engine covers them all.
Models are typed functions. A detector goes from video to detections. If one model's output matches the next one's input they chain. If it does not, the pipeline does not compile.
So the next job is a small layer on top, not a new project.
The engine ships as apps.
Same engine, a different front door for each job.
- Redactio Video redaction. Flagship
- Claritas Footage restoration. Beta
- Kinetix Motion classification. Beta
- Composer Brand video at scale. Coming soon
- Verbatim Transcription and audio PII. Planned
- Ledger Audit and chain of custody. Planned
Alpha access is by conversation, not by form.
Three claims, all checkable.
In the EU
Storage, detection, rendering and export are built to run in eu-central. Footage and audit records never leave the EU.
Never trains a model
Your footage is not used to train or fine-tune models. Contractual, not a settings toggle.
A sealed receipt
Every export carries the SHA-256 of input and output, the sealed plan, and the model weights by hash. An offline verifier re-checks it with no account.
We do not decorate with certifications we have not earned.
How many versions of the same video does your team make in a month?
Bring us a job. Fifteen minutes is usually enough to say whether the engine can do it.