Fix poor lighting in product photos and keep the true colour

A photo taken in poor light looks cheap, even when the product is good. Worse, if the colour on screen is not the colour in the box, you get returns and angry messages. Product photos are not portraits: the job is not to look pretty, it is to show the real thing.

This page puts the options in order from cheapest to most complex, explains where AI relighting fits, and gives you a colour check and a short test with pass and fail rules. I use Higgsfield's Relight app as the AI example, and I say plainly when a simpler route wins.

This page may contain affiliate links from which Blur Studio earns commissions.

Why poor light costs you

Buyers cannot touch the item. The photo does the touching. Dark, grey or tinted images read as low quality, and a colour that is off can lead straight to a return. I could not find an independent study that measures AI relighting against returns. Figures such as "30 percent of returns come from image expectations" appear on vendor blogs without sources, so I do not use them.

Here is an illustrative example with made-up numbers. A product sells 200 units a month at $40. If one extra return per 100 sales is blamed on colour mismatch, that is 2 returns a month, about $80 of revenue plus shipping and handling. If you sell a colour-critical item, a month of mismatch returns can cost more than a reshoot. For a plain black phone case, it will not. Use your own numbers.

Diagnose the problem first

CapCut, a video and photo editing app, has a useful checklist for this. Work out which one you have, because each has a different fix.

What you see	Likely cause	Best fix
Whole photo dark	Underexposed	Retake with more light, or raise exposure
Washed out, details lost	Overexposed	Retake, or lower highlights
Hard black shadows on one side	Harsh single light	Retake with a white card to fill, or soften
Orange or yellow tint	Indoor bulbs	Fix white balance, or retake by a window
Blue tint	Shade or screen light	Fix white balance
Warm patch on one side, cool patch on the other	Mixed light, such as a window plus a bulb	Switch the bulb off and use the window alone, or retake
Subject dark with bright background	Backlit	Retake, or a directional relight

Cheapest fix: retake near a window, with a grey card

If the product is in your hands, this wins. Put it beside a window, out of direct sun, with a white sheet or card on the shadow side to bounce light back. A cheap ring light is another route, and your local shop will quote the price.

One trick saves a lot of trouble later. Shoot one extra frame with the product beside a neutral grey card or a plain white sheet, even if the light is poor. That frame lets any editor set white balance numerically, and it gives you a colour anchor to judge any AI result against.

Two more cheap fixes from photography guides such as Digital Photography School: use a reflector (any white or silver board) to push light into the shadow side, and move the product to the light instead of trying to repair a bad spot afterwards. If you must use a flash, a gentle fill flash lifts shadows, but it can cause glare on shiny packaging, so check the result.

If your phone or camera can shoot RAW (a file that keeps more light data than a normal JPEG), use it. CapCut's guide says RAW files leave you far more room to recover dark and bright areas afterwards.

Free slider editing: what it fixes and what it cannot

Sliders adjust brightness, contrast, shadows, highlights and warmth on the pixels you already have. They do not invent anything. That makes them the safest route for colour. Cloudinary, a media software company, offers a free browser photo lighting editor with brightness, contrast, midtone gamma and warmth controls, with no watermark, and it describes the result as real lighting corrections rather than full AI scene regeneration. Phone editors and Canva, an online design tool, do similar work. Canva's pricing page, read in a United Arab Emirates view on 3 November 2026, showed Free at AED 0 and Pro at AED 440 per person per year, with AI tools sharing a monthly allowance. Prices vary by country.

Sliders cannot fix a subject that is dark while the background is bright, or flat light that needs a side light to show shape. For that, you need to move the light, not just brighten.

One-click AI "enhance" buttons sit in between. CapCut and Photoroom, a product-photo editing app, offer them. They are quick for simple brightening. Check the colour afterwards in the same way as any other edit.

Generative relight: how it works, and where Higgsfield fits

Relighting asks software to redraw the picture as if the light came from somewhere else. You choose a direction, how soft or hard the light is, its brightness and its colour. The picture is generated again, so colours can change. That is the risk.

Higgsfield's Relight is one of the small image apps on its Apps page, described there as a way to adjust light position, colour and brightness. PetaPixel, a photography news site, reported it in January 2026: you can change the angle of the light source, colour temperature and brightness, with a 3D directional pad, a hard or soft toggle and a colour input, aimed at product photographers and also at selfies. PetaPixel mostly repeats Higgsfield's own wording and does not describe a trial of the tool. AI Tool Curator, a review blog, describes the same controls as direction presets, soft or hard light, brightness and a colour value. Neither source reports how well it keeps product colour, so test five images before running fifty.

What I can and cannot tell you, as of November 2026:

• The Apps page marks Relight with a PRO badge. Higgsfield does not explain what the badge means, so check which plan you need on the live pricing page and our pricing explainer.
• The credit cost of Relight is not published. A credit is the platform's unit of spend. The cost shows on the button before you run it. See how to spend credits wisely.
• Claims that relighting "changes the light, not the product colour" come from vendors. I found no independent test, so you must run the colour check below.
• Free accounts put a Higgsfield watermark on every generation. Do not test or publish client images from a free account. Higgsfield's pricing table marks commercial use "Not included" on Free, while its help center says outputs are not restricted to paid plans, so use a paid plan for client work.

For the wider tool picture, see Higgsfield upscale, relight and edit tools and best AI tools for product photography.

The colour check: compare with the real product

This takes about three minutes and catches the failures that cost you money.

1. Put the real product next to your screen, in daylight, with a grey card or white sheet if you have one.
2. Pick three colour patches on the product: for example the main body, an accent and a shadow area.
3. Look at those patches in the original, in the slider edit and in the relit version.
4. Reject any version where a patch visibly shifts compared with the real item.

Check texture, labels and logos too. Nightjar, an AI product photography company, advises checking colour, materials, text, logos, seams and reflections against the physical item before publishing. Fine print and logos are where generated light tends to smudge. Also zoom to 100 percent and check the export size, because some relight tools limit the output size, and a small file will look soft on a product page.

Where the image will be shown

Marketplace rules are strict. Photoroom's summary of Amazon's main image rules, which I have not checked against Amazon's own page, says the main image should show the product only on a pure white background, fill about 85 percent of the frame and be realistic, sharp and colour accurate, without heavy filters. So a creative relight belongs on secondary lifestyle images, not the main one. Read your marketplace's own guide. My article on marketplace product images covers what gets rejected.

A seven-day test with pass and fail rules

This is a plan you can run. The thresholds are my suggestions.

1. Day 1: pick ten poorly lit product photos with the real products on hand.
2. Day 2: edit five with sliders only. Edit five with Relight. Log the time for each.
3. Day 3: run the colour check on all ten.
4. Day 4: ask two people who know the real product to vote "looks like the real thing", yes or no, on each image, without telling them which method was used.
5. Day 5: for any method that passed, test five more.
6. Days 6 and 7: decide and write the settings down.

Keep a method only if at least nine of ten pass the colour check and the edit takes under five minutes per image. Drop it if two or more images show a colour shift. If sliders win, you have saved the subscription.

Batch rules: settings log and sample first

If you light a whole product line, record the settings you used on the hero shot: direction, soft or hard, brightness and colour value. Reuse them on the rest. Test five images before running fifty, because no source I found shows how consistent the results are across a batch. Keep the original of every image, and write down which method made each file in your product sheet. For a whole shoot, see batch retouching options and natural-skin checks.

Products you should never relight

• Items where colour is the purchase decision: paint, cosmetics, fabric, jewellery with stones, food.
• Images used as evidence: damage, condition, used goods.
• Anything where the colour on delivery will be noticeably different from the photo.

Who should skip relight and use a photographer

• Shops whose products are colour-critical. A light tent or a photographer gives consistent colour across a catalogue.
• Anyone who can retake the photo. Free and exact.
• People who only need brighter exposure. Sliders do it without generating new pixels.
• Skin-tone work such as portraits, which belongs to a different tool and has consent rules. See team headshots without booking a photographer.

Also note that output is not guaranteed to be unique, and that if your photos are confidential you should read Higgsfield's terms on training. See is Higgsfield safe and legit.

How I'd approach this

I would retake first, with a window and a grey card frame. I would do a slider edit second. I would use a generative relight only on photos where the problem is the direction of the light, and only after the colour check passes. If you want to try the Relight app, open a Higgsfield account on a paid plan and read the cost on the button before your first run. For new views of the same product rather than new light, see getting many angles from one hero product photo.

If you handle hundreds of photos, the repeatable part is the checklist, the settings log and a place where the right file is stored. Blur Studio builds that kind of creative workflow, approval portal or asset library around the tools you pick.