The Science

No magic. Just a stack of science you can read.

The clever bits behind LightVision aren't magic — they're proven ideas from machine learning, most of it running privately on your own machine. Here's what's actually going on, in plain English.

Every feature here is a well-understood idea — pointed at the one thing rivals can't copy: you. LightVision learns your taste from the frames you keep and your look from the edits you ship, all on your own computer.

On-device

Culling, grading and search run on your machine — privacy by design, not a promise.

Glass-box

Every call breaks down into reasons you can see, never one mystery verdict.

Reversible

Edits are sliders on a non-destructive stack — anything the AI does, you can undo.

It learns you

Seeded from your own keepers and edits, so it converges on your eye over time.

Act I

Seeing

First it has to understand what it is looking at — quality, focus, content and who is in the frame.

Seven verdicts, not one

Every frame gets seven readable scores — not one mystery number.

Photo being graded by LightVision Studio

Instead of a single thumbs-up, LightVision scores each photo on clarity, focus, open eyes, emotion, composition, moment and exposure, then sorts it into Pick, Alternate or Review. You always see why a frame made the cut — and you can overrule it.

The science · Multi-attribute scoring

✓Hours of culling become a quick review of reasons you can trust.

Sharp is not the same as in focus

It catches the shots where the eye just missed focus.

A photo can look crisp in a thumbnail yet be soft on the eye — the one place that has to be sharp. LightVision measures exactly where the focus landed and flags the near-misses, so you never keep the wrong frame.

The science · Edge-energy focus measurement

✓No more finding a soft-eyed 'keeper' at full size.

Find any moment in plain words

Search your shoot the way you'd describe it.

Type 'first kiss' or 'confetti toss' and the matching frames surface instantly — no tags, no filenames, no endless scrolling. It understands what a photo is about, not just what it's called.

The science · Vision-language search

✓Pull any moment from thousands of frames in seconds.

Everyone, recognised across the day

The same person, gathered from every corner of the shoot.

LightVision recognises each face and groups every photo of the same person — shot hours apart, in totally different light. Duplicates of the bride collapse to one decision, while two different guests never get merged.

The science · Face clustering

✓One clean decision per person, not per frame.

Act II

Learning your eye

The part rivals can't copy: it stops judging photos in general and starts learning how you judge them.

It learns your taste, not rules

Every keep quietly teaches it what you would choose.

You never write rules — you just cull. From the frames you keep over the ones you cut, LightVision learns your eye and floats the photos you'd pick to the top of the next shoot.

The science · Preference learning

✓The more you use it, the more it culls like you.

It learns your look, not a preset

A first-pass edit in your style — tuned to each photo.

A preset slaps the same edit on everything. LightVision learns how you edit and predicts the right adjustment for each frame, so a bright outdoor shot and a candle-lit one each get the proper touch in your voice.

The science · Learned edit model

✓New imports arrive most of the way to your final look.

Bootstrapped from your finished work

Already shaped like you, from day one.

Point LightVision at galleries you've already delivered and it reads the look you've spent years refining — shrugging off the odd dramatic edit and learning your typical style per scene. Your very first import already sounds like you.

The science · Robust statistics

✓No cold start — it grades and edits in your voice immediately.

Act III

Acting with precision

Now it changes pixels — exactly, locally, and always reversibly.

Select anything by name

Name what you want changed — and it's selected.

The background selected just by naming it

Type the thing you want to adjust — the background, the sky, a dress, a face — and LightVision finds it and draws a clean, soft-edged selection for you. Local edits that used to mean minutes of careful masking now take seconds.

The science · Language-guided selection

✓Pro-level local edits without the tedious masking.

Edit by talking to it

Plain words become real, adjustable sliders.

Say 'warm it up and lift the shadows' and LightVision moves the actual sliders — the same ones you'd drag by hand. Nothing is baked in: tweak it, stack on more, or undo it entirely.

The science · Language to controls

✓Describe the look you want and stay in full control.

Repair that still looks like them

Cleaner portraits that never become a stranger.

A portrait restored while staying true to the person

Restoration sharpens eyes, clears noise and tidies skin while staying locked to the real person — so your subject looks like themselves, only better. No plastic, generic 'AI face.' Drag the slider to compare.

The science · Identity-anchored restore

✓Flawless portraits that are unmistakably your subject.

Act IV

The foundation

All of it rests on one choice: the intelligence runs where your photos already live.

It all runs on your machine

Private by design — and still fast.

Culling, grading, style learning and search all run on your own computer. Your shoot never has to leave your machine — privacy isn't a promise here, it's how it's built — and it's quick enough for a real deadline.

The science · On-device AI

✓Your clients' photos stay yours, full stop.

The philosophy

The most useful thing we can tell a pro is not "the AI handled it" — but exactly how it works, where it's confident, and where the dial is that puts you back in control.

See the science do your culling.

Start a free trial and watch every idea on this page run on your next real shoot.

Start free trial Calculate your time saved