Most product photo lighting problems are fixable after the shoot. Exposure, colour cast, an uneven white background and a shadow falling the wrong way are all recoverable from a sharp original. One failure is not. A highlight blown to pure white holds no data, so nothing can bring the detail back. Sort your images by that line before you book a reshoot.
You have 40 product photos from a shoot that went fine. On the contact sheet they look consistent. On the listing page, six of them read grey where the background should be white, three have a warm cast the others do not, and one hero shot has a hot spot across the lid where the logo sits.
Nobody wants to hear that a reshoot is the answer, because a reshoot means the studio, the stylist, and the products shipped back and forth. Often it is not the answer.
This covers which lighting failures recover and which do not, why a rejected white background is almost always a lighting problem, the accuracy line you should not cross when correcting colour, and the setup that stops the whole category of problem at source.
Key Takeaways
- Three of the four common lighting failures are recoverable. Underexposure, colour cast and uneven background all survive correction if the original is sharp and the highlights are intact.
- Clipped highlights are the exception and they are permanent. Once a region reads pure white with no detail, there is no information left to recover, and no tool invents the missing texture honestly.
- A rejected “white background” is a lighting fault, not a background fault. Amazon requires pure white at RGB 255, 255, 255 with the product filling at least 85% of the frame (sellercentral.amazon.com, August 2026). Uneven light across a white sweep renders the corners grey, so the sweep is white and the photo is not.
- Correcting light is fine. Correcting the product is not. Push a colour cast out of the frame and you fix the photo. Push the fabric two shades warmer and you have changed what the buyer thinks they are buying.
- One light direction across the whole set matters more than any single image. Shadows are the only cue telling a viewer where the light came from, and mismatched directions make a collection page look assembled from three shoots.
- AI relighting works on the same physics as the shoot. It rebuilds the light on a product it can already see. It cannot reveal detail the original never captured.
Which lighting problems can be fixed after the shoot
A digital image is a record of how much light hit each pixel. Correction works when the record still holds usable values in the area you want to change. It fails when the values are already at the top or bottom of the range, because everything in that region has collapsed to a single number. Underexposed shadows usually hold recoverable data. Blown highlights usually do not.
That single fact sorts every lighting problem you will meet into fixable and not.
The four failures, ranked by how recoverable they are
| Failure | What it looks like | Recoverable? | What it costs to fix |
|---|---|---|---|
| Underexposure | The whole frame reads dark and muddy | Yes, usually fully | Minutes, plus some noise in the shadows |
| Colour cast | Whites read warm, blue or green | Yes, if the cast is even across the frame | Minutes, one white balance correction |
| Uneven background | White sweep goes grey toward the corners | Yes | Minutes, or a cutout and a clean fill |
| Clipped highlights | A region is pure white with no texture | No | A reshoot of that product |
Underexposure is the safest mistake to make. Shooting a third of a stop under leaves headroom in the highlights and puts the risk in the shadows, where the sensor still holds detail you can lift.
Colour cast is the second safest, with one condition: the cast has to be even. A single warm source across the whole frame corrects with one adjustment. Two sources at different colour temperatures, say a window and a tungsten lamp, leave a split you have to fix in separate areas of the same image.
An uneven background is a geometry problem wearing a lighting costume. The sweep is white. It reads grey in the corners because less light reached them. That is fixable, and it is also the one most likely to get a listing suppressed before you notice.
Clipped highlights are the wall. A specular hot spot on a glossy lid, a bright edge on a chrome fitting, a window reflection on packaging film. The pixels are at maximum. There is nothing under them.
Why a rejected white background is a lighting problem
Amazon’s main image rule is specific: pure white background at RGB 255, 255, 255, product filling at least 85% or more of the frame, no text, logos, borders or watermarks, and the image must not be blurry or pixelated (sellercentral.amazon.com, August 2026). Images run between 500 and 10,000 pixels on the longest side, and 1,000 pixels or more enables zoom on the detail page.
Sellers read “white background” and buy a white sweep. Then the listing gets flagged anyway.
The reason is that RGB 255, 255, 255 describes the pixels in the file, not the paper in the studio. White paper photographed with a light source aimed at the product falls off toward the edges of the frame. The middle reads 255. The corners read 240, or 232. To the eye on a bright monitor it looks white. To an automated check it is not white.
Two ways to solve it, and they are not equivalent:
- Light the background separately from the product. A second source aimed at the sweep, roughly one stop brighter than the key on the product, puts the whole background at maximum without spilling light back onto the front of the item.
- Cut the product out and place it on a true white fill. Faster, and it is what most catalogue pipelines do at volume. This is the job background removal does, and the part that matters is what happens after the cutout: edge quality on hair, mesh, glass and anything semi-transparent.
Route two is cheaper and route one is more honest to the original object. Pick per product. Anything with a soft or translucent edge is better lit properly than cut out.
For the full slot-by-slot picture across channels, the ecommerce product photography specs piece covers what each marketplace actually shows.
The line between fixing the light and changing the product
This is where correction turns into a compliance question.
Removing a green cast from a room lit by fluorescent tubes returns the image toward what the product looks like in neutral light. That is a fix. Warming a beige sofa until it photographs as camel because camel is selling better is not a fix, it is a different product.
The test is simple. Ask whether the correction moves the image toward the object in the room, or away from it. Colour, finish, texture and scale are the four attributes a buyer claims they were shown. Everything else is presentation.
Colour is the one that generates returns. A screen never renders fabric and wood tone exactly, so some drift is unavoidable, and that is the argument for controlling the input rather than correcting hard at the end. Shoot a grey card or a colour reference in the first frame of every setup. It costs one shutter actuation and it turns white balance from a judgment call into a measurement.
There is more on where that line sits in what to check before you ship a product photo.
How AI relighting works, and where it stops
AI relighting rebuilds the lighting on a product the model can already see. It reads the shape and material from the existing pixels, estimates where the original light came from, then re-renders the scene as though the light had been somewhere else. DesignerBox runs this as Spotlight, and as a saved cinematic relight workflow when the same treatment has to run across a batch.
Three things it does well:
- Flat lighting into directional lighting. A product shot under a single overhead source has no form. Relighting adds a key direction and gives the object edges again.
- Matching a set. Forty products shot across three days under drifting daylight can be pulled to one light direction, which is the thing that makes a collection page look deliberate.
- Restyling without a reshoot. The same packshot relit warm for an autumn campaign and neutral for the standard listing.
Two things it does not do:
- Recover clipped detail. If the logo on the lid is a white blob, relighting produces a well-lit white blob. The information is gone.
- Fix softness. Lighting and focus are separate problems. A blurry image relit is a blurry image. That is a different repair, covered in low resolution product images.
The honest framing is that relighting is a lighting tool, not a rescue tool. It works on the same physics the shoot did.
A setup that stops the problem happening
Fixing lighting is cheaper than reshooting. Not needing to fix it is cheaper than both.
Use one light direction for the entire catalogue and write it down. Not “soft light from the left”. The actual position: source height, distance from the product, angle relative to the camera, and whether a bounce card sits opposite. A phone photo of your own rig is a legitimate production document, and it is the only thing that makes the next drop match this one.
Make the key source large relative to the product. A small source produces hard-edged shadows and tight specular hot spots, which is how highlights get clipped. A large source relative to the subject wraps the light and spreads the highlight over a wider area, so it stays inside the range the sensor can hold.
Add a second source for the background, not for the product. Separating the two is what gives you a true white without flattening the item.
Expose for the highlights. Protect the top of the range and lift the shadows later. The classic three-light arrangement of key, fill and rim is documented at three-point lighting if you want the roles named.
Shoot a reference frame first. Grey card, colour checker, or at minimum a known-white object in the corner of the first shot of each setup.
The same discipline is what makes overhead work hold together across a set, which is why flat lay photography comes down to camera geometry and one light direction rather than styling.
What to check before it ships
Run this on every batch:
- Sample the background corners, not the centre. Pick the four corners of the white area and read the values. If any reads below 250, the light was uneven.
- Check the highlights at 100%. Zoom into the brightest region of the product. If there is no texture, it is clipped, and that image is a reshoot.
- Compare whites across the set. Open six images side by side. A cast that is invisible alone is obvious in a row.
- Check the shadow direction. All shadows should fall the same way across the set.
- Check small type at 100%. Ingredient lists, model numbers and care labels are where softness and hot spots hide.
- Confirm the frame fill. Amazon wants 85% or more on the main image, so a product floating in white space fails on framing even when the light is perfect.
That last one is a framing job rather than a lighting one, and the complete product photography shot list covers which angles need to exist before you worry about how any of them are lit.
FAQ
Can you fix bad lighting in a product photo?
Usually yes. Underexposure, an even colour cast and an uneven white background are all correctable from a sharp original. The exception is a clipped highlight, where a region has recorded pure white with no detail underneath. That information does not exist in the file, so no correction and no AI relighting can return it. Those images need a reshoot.
Why does my white background keep getting rejected?
Because the paper is white and the pixels are not. Amazon requires RGB 255, 255, 255 on the main image (sellercentral.amazon.com, August 2026), and light falling off toward the corners of a sweep leaves them reading grey. Light the background separately from the product, roughly one stop brighter than the key, or cut the product out and place it on a true white fill.
What is the best lighting setup for product photography?
One large soft key source positioned consistently, a bounce card or fill opposite it to open the shadows, and a separate source on the background if you need a pure white. Size matters more than count: a large source relative to the product spreads highlights and reduces the chance of clipping. Record the positions so the next batch matches.
Does AI relighting change the product?
It should not, and that is the thing to check. Relighting rebuilds the light, so shape, finish and texture stay as photographed. It becomes a problem when the correction shifts colour far enough that the fabric or finish no longer matches the physical item, because colour is one of the attributes a buyer claims they were shown.
Should I underexpose product photos on purpose?
Slightly, yes. A third of a stop under protects the highlights, which is the one part of the range you cannot recover. Shadows hold usable data much further down. This matters most on glossy, chrome and glass products where a specular hot spot clips before the rest of the frame looks bright.
How do I keep lighting consistent across a whole catalogue?
Fix the light positions physically and record them, then treat any drift as a correction pass rather than a per-image judgment. When products are shot across multiple days or sourced from suppliers with their own photos, pull the whole set to one light direction in a single pass instead of editing image by image. A saved workflow does this repeatably, which is the point of running it as a batch relight rather than one file at a time.
Can I relight a supplier’s photo instead of reshooting the product?
Yes, if the supplier image is sharp and correctly framed. Relighting can bring a mixed set of supplier photos to one consistent look, which is the common case for a catalogue you did not shoot. It cannot fix softness, low resolution, or a product that is cropped wrong. Check those three first, because they decide whether the image is worth the pass at all.
Fix the light, then keep it fixed. DesignerBox relights product shots to a set standard and saves the treatment as a workflow the whole catalogue can run through, so every drop comes out matching the last one.
Amazon image requirements verified at sellercentral.amazon.com as of August 2026. Lighting physics and correction limits are properties of digital capture, not vendor claims. Individual results vary.
