PngText

AI image Generator editing

AI Relight

Brightening a photo and relighting it are not the same operation, though they are sold as if they were. One raises every tone together and flattens the picture; the other decides where the light is coming from, how hard it is, what color it is, and what is standing in its way. The second is what an ai relight pass actually does, which is why a good result looks like a different time of day rather than a brighter version of the same one.

You moved the light, you did not turn it up

AI Relight

A photograph carries a record of the light that made it: shadows point somewhere, highlights sit on one side, and the softness of the edge between them says how big the source was. An ai relight pass rewrites that record. It can make a lamp-lit room read as window light or turn a flat overcast shot into something with a direction to it, and it does that without changing who or what is in the frame.

Light direction rebuilt rather than brightness raised
Shadow edges matched to a believable source size
Color of the light separated from the color of the objects
Areas opened by the new light filled in rather than left blank

Relight the photo

A flat portrait beside the same frame with a new source placed to one side and shadows falling away from it

AI Relight

Four properties, and only one of them is amount

What is actually being decided when the light is rebuilt.

Light in a picture is described by direction, by how hard or soft its edges are, by its color, and by what it is blocked by. Amount is not on that list, and a tool that only moves amount is a brightness slider with better manners. Everything below assumes you are using an ai relight pass to change one of the other four, and that the picture will be judged on whether the new light agrees with the shadows that are already there.

The same portrait brightened globally beside the same portrait relit from one side with part of the face falling into shadow

AI Relight

Raising brightness lifts everything, including the light

A brightness adjustment multiplies every tone by the same factor, so the dark corners come up and the bright areas clip at the same time. That is why a brightened photo looks flat rather than better lit: the range between the darkest and lightest part has not changed, it has just moved. Relighting is not a global multiply. It assigns a source, gives that source a position, and then darkens and lightens different surfaces by different amounts according to which way they face. Anyone who wants an ai relight result to look like real light has to accept that some parts of the frame will get darker than they started, which no brightness control will ever do.

A portrait with shadow direction and highlight side marked to show the light the camera already recorded

AI Relight

The photograph already contains a plan of its own light

Before anything is changed, the original tells you where the light came from. Shadows run in one direction, highlights gather on the surfaces that face the source, and the width of the zone between lit and unlit says how large that source was relative to the subject. Reading those three things is the whole preparation for an ai relight pass, because the result is judged against them. A new light that contradicts the old one, brighter on the side where the shadows already fall, reads as wrong immediately even when nobody can name why.

Two casts of the same shadow, one with a hard narrow edge and one with a wide soft transition, from sources of different size

AI Relight

Hardness is a question of size, not of strength

A small source close to the subject throws a hard edge with a narrow transition, the way a bare bulb or a phone flash does. A large source, a window, an overcast sky, a bounce card, spreads the transition over a wide band and the result looks gentle. This is independent of intensity: a dim window is still a soft light, and a bright bulb is still hard. So an ai relight that produces soft edges in a scene shot under direct midday sun has invented a different day. Matching the transition width to something the scene could plausibly contain is what keeps the change believable.

A dark area of a photograph lifted to show reconstructed texture where the original recorded almost nothing

AI Relight

Opening a shadow means inventing what was under it

Relighting implies lifting areas that were dark, and those areas hold very little recorded detail. A shadow under a chin, inside a doorway, or across half a room is mostly compressed noise, and once the new light brightens it the tool has to supply texture, tone, and edge that the sensor never captured. That is the honest limit of an ai relight result: the parts of the frame that were unlit are reconstructed rather than revealed. It works best when the surface is simple and predictable, a wall, a floor, a plain fabric, and it shows most when the dark area contained something specific, like lettering or a pattern.

A cheek and a glass bottle with the size of the reflected highlight indicating how large the light source is

AI Relight

Skin and glossy objects give the new light away

A specular highlight carries the shape of its source. On skin, on a bottle, on a car panel, or on wet fruit, the bright patch is a small portrait of the lamp that made it, and its width and softness are readable. Getting that wrong is why a relit face can look like a mannequin: the texture is right but the shine is the wrong size. Paying attention to how broad the highlight should be on each rounded surface is what separates a convincing ai relight pass from one that merely looks lit.

A neutral object under warm light and under cool light with its own color unchanged in both

AI Relight

Color belongs to the light, and objects keep their own

Warm light and warm paint are two different things, and a relight has to move one without the other. Raising the color temperature of the source should tint the parts of the scene the light reaches and leave the object's own hue intact underneath, so a gray wall becomes a warm gray and a blue shirt stays blue rather than turning teal. When the two get mixed, the picture looks color graded instead of relit. Keeping the source color and the surface color separate is what an ai relight pass is being asked to do, and it is the reason the operation cannot be replaced by a filter.

A backlit silhouette at sunset beside the same frame relit flat, with the atmosphere gone

AI Relight

Some photographs should keep the light they were taken in

Strong existing light is often the subject. A figure against a low sun, a room lit by one window at dusk, a stage lit in magenta and green: relighting these does not improve them, it removes the reason they are interesting. An ai relight pass is at its most useful on pictures where the light was merely inadequate, a flat indoor shot, a hard flash, a dull overcast day, and at its most destructive on pictures where the light was the point. Deciding which kind of photograph you have comes before deciding to relight it.

AI Relight

What an ai relight pass covers

The properties that change, and the ones that do not.

Source given a position

The light is placed somewhere in relation to the subject, so the surfaces that face it brighten and the ones turned away fall back, instead of every tone moving together.

Edge width matched to a real source

The transition between lit and unlit is set to a width that suits a plausible light, from the tight boundary of a bare bulb to the broad falloff of a window.

Color carried by the light alone

Warmth or coolness is applied to what the source reaches, with the objects keeping their own hue, so the frame still reads as a photograph rather than a grade.

Unlit areas reconstructed, not revealed

Where the new light opens a shadow, the missing texture is built from the surrounding material. Simple surfaces come back well, and busy ones show the guesswork.

Faces and identity protected

Features, proportions, hair, and clothing stay where they were, since an ai relight pass is about where the light sits rather than about who is standing in it.

Strength kept honest

The change can be dialed from a gentle correction of a dull frame to a full reset of the lighting, and the point where it stops looking photographed is easy to find.

AI Relight

Relighting a photograph

Read the existing light first, then decide what an ai relight pass should replace it with.

Read the light the photo already has

Note which way the shadows run, which surfaces are bright, and how wide the boundary between lit and unlit is. That record is what the new light has to either agree with or deliberately replace, and reading it takes less time than fixing a contradiction later.

Decide what should change and what should stay

Usually one property is the problem: a flat frame needs a direction, a harsh flash needs a softer edge, a dull indoor shot needs warmth. Naming the single property keeps an ai relight pass from quietly rewriting the other three along with it.

State the source and keep the subject fixed

Describe where the light should come from, how large it should behave, and what color it should be, then say explicitly that the subject and the scene stay as they are. Both halves are needed for a usable ai relight result.

Check the shadows and the shine at full size

Look at the transition along the shadow edges, then at the small highlights on skin, glass, and metal. Those two places are where a wrong source size shows, and both are invisible in a thumbnail.

AI Relight

Questions about relighting a photo

Brightness, shadow recovery, skin, and when an ai relight result should be avoided.

How is this different from making a photo brighter?

Brightness multiplies every tone by one factor, so highlights clip while the shadows only creep up and the picture flattens. An ai relight pass chooses a source, gives it a direction and a size, and then changes different surfaces by different amounts, which means some areas end up darker than they began.

Will the shadows still look natural?

They should be the first thing you check. Shadows have to fall away from the new source and their edges have to be the width a source of that size would produce. Where the two disagree the picture reads as edited even to someone who could not explain why.

What happens to the detail that was hidden in shadow?

It gets rebuilt rather than recovered. Areas that were nearly black hold very little recorded information, so brightening them means constructing texture and tone that fit their surroundings. Plain surfaces come back convincingly and detailed ones, lettering, prints, fine patterns, are where the reconstruction is visible.

Can it make a dull indoor picture look like it was shot by a window?

That is one of the better uses of it. A flat frame with no clear direction is exactly what an ai relight pass can improve, because there is little existing light to contradict. Aim the new source from one side, set its edge somewhere between hard and soft, and give it the neutral color of daylight.

Why does a relit face sometimes look like a mannequin?

Because the highlights are the wrong size. The bright patch on a cheek or a nose is a small image of whatever lit it, and if its width does not match a source that could exist in that scene, the skin reads as plastic even when the shading itself is correct.

Should every poorly lit photo be relit?

No. Pictures whose light is the subject, a backlit silhouette, a room at dusk with one window lit, a stage in colored light, lose their point when the light is rebuilt. Keep an ai relight pass for frames where the light was merely weak or directionless.

Does the original file get changed?

Only if you overwrite it. Keep the source photo and work on a copy, since the light the camera recorded is the one reference that lets you judge whether the new version is better or simply different.