How to Create Product Colour Variants From One Photo With AI
A practical workflow for turning one product photo into multiple colour variants with AI, while preserving texture, product geometry and accurate ecommerce expectations.
If you sell the same product in several colours, photographing every variant can turn one simple catalogue update into another full shoot. AI recolouring can reduce that work by starting from one approved product photograph and changing only the colour while trying to preserve texture, lighting and shape.
The workflow is useful, but there is an important ecommerce distinction. A generated colour variant is not automatically proof of what the physical variant looks like. If the new colour is already in stock, compare the result with the real sample. If the colour does not exist yet, treat the image as a concept or testing asset rather than a factual product photograph.
Photoroom's current Recolor tool is a useful example because it is built specifically for changing product colours. It accepts a source image, a target colour and an optional prompt describing exactly which part should change. Photoroom says the tool is designed to preserve texture and detail, although it currently works best with single-colour products and uses AI credits.
Step 1: Choose the best source photo, not just the prettiest one
Start with a genuine photograph where the product shape and surface are easy to read. Good edge definition matters because the recolour needs to understand where the product ends. Visible texture matters because folds, grain, stitching, engraving and reflections are what stop the new colour from looking like flat paint.
For clothing, use a photo where fabric folds and seams are clear. For packaging, make sure logos and labels are sharp. For jewellery, use a clean image where the clasp, chain, engraving, stones and polished or brushed finish are visible. Avoid starting from an image that is already overexposed or heavily colour-cast, because the AI then has a weaker reference for the material itself.
Step 2: Change one colour region at a time
In Photoroom Recolor, upload the source image, choose the target colour, then use the prompt field to describe which area should change. Be specific. Instead of saying “make this product blue,” write “change only the body of the bottle to navy blue; keep the cap, label, logo and background unchanged.”
This matters because many products contain more than one meaningful colour. A trainer may have white soles, black laces and a coloured upper. A bracelet may have a silver chain with a coloured charm. A cosmetics bottle may have a metallic cap and printed label. Asking for one broad recolour gives the model more freedom to change parts that should stay fixed.
Photoroom's current documentation explicitly says Recolor is strongest on single-colour products and may work with white, grey or black trims. More complex colour schemes still need closer review or separate edits.
Step 3: Use a real colour reference when accuracy matters
A colour picker or hex value can make the instruction more precise, but screen colour is not the same thing as physical material colour. The same RGB value can look different on cotton, glossy plastic, brushed steel or translucent glass because the surface changes how light is reflected.
If the physical variant already exists, photograph a small real sample under neutral light and keep it beside the generated image. The AI output should move toward the sample, not toward whichever version looks most attractive on screen. This is especially important for ecommerce because customers are buying the material, not the hexadecimal colour value.
Step 4: Check whether the texture survived
After generation, zoom in before judging the new colour. Look at folds, stitching, embossing, surface grain, highlights and small shadows. A poor recolour often looks acceptable as a thumbnail but turns material texture into a smooth colour overlay when viewed closely.
Photoroom positions Recolor around preserving texture, detail and natural shading rather than applying a flat tint. That is the right goal, but it is still a generated result. Compare it with the source and reject any version where seams disappear, text changes, edges soften or the material starts behaving like a different surface.
Step 5: Keep the lighting consistent across the colour set
One advantage of starting from the same source photo is consistency. The camera angle, crop, product size and general light direction can remain the same across the full colour range. That can make a product grid feel much cleaner than several shoots made on different days.
Do not undo that benefit by asking AI to change the background, camera angle and colour in the same step. Create the colour variants first. Once the approved set is consistent, adapt the background or campaign styling separately if needed. Smaller transformations are easier to compare and easier to reject when something goes wrong.
Jewellery needs a different definition of “colour”
Gold, silver, rose gold and black jewellery finishes are not flat colours. The appearance comes from the base tone plus reflections, highlight colour, roughness and the way the plated or polished surface reacts to the environment.
That is why simply turning a silver bracelet yellow can produce fake-looking gold. A convincing metal variation needs the material response to change with the colour while the underlying jewellery geometry stays fixed. Check polished edges, dark reflections, bright highlights, engraving, clasp shape, chain links and stones separately.
For jewellery sellers using Lustra Studio, colour variation is better treated as a product-specific workflow rather than a generic tint. Multiple real jewellery references can help anchor the structure while the finish changes. Even then, a genuine sample of each metal finish remains the best final reference when the generated image will represent a stocked SKU.
Step 6: Separate catalogue variants from colour testing
AI recolouring has two different uses, and they should not be confused. The first is production: you already manufacture the red, blue and black versions and want consistent images of all three. In that case, verify each generated image against its physical variant before publishing.
The second is experimentation: you are considering a new colour and want to see how it might look before ordering samples. Photoroom itself promotes Recolor for testing colourways before committing to production. That can be useful for internal decisions, ad concept tests or early feedback, but the generated image should not imply that a nonexistent sample has already been manufactured and accurately photographed.
A simple approval checklist
Before using a recoloured image commercially, compare it with the source and check five things: the product shape is unchanged, fixed details such as logos or engravings are unchanged, texture is still visible, shadows and highlights make sense for the new material, and the displayed colour resembles the real variant where one exists.
If the colour is correct but a logo changed, the image fails. If the product geometry is perfect but the rose-gold finish looks like orange paint, it fails. Recolouring only saves photography work when the result still represents what the customer will receive.
The takeaway
AI can turn one strong product photograph into a consistent set of colour variants without repeating the entire shoot. The safest workflow is narrow: start from a real source, recolour one region at a time, use precise colour guidance, preserve the same lighting and framing, then verify texture, fixed details and material behaviour before publishing.
For simple single-colour products, current tools such as Photoroom Recolor make that process straightforward. For reflective jewellery and multi-material products, expect more review because changing colour also changes how the material should interact with light.
The useful promise is not “one photo can become any product.” It is “one accurate photo can become a controlled starting point for variants, as long as the real SKU remains the final authority.”
Sources: Photoroom Help Center, “How to use Recolor,” updated 29 August 2026; Photoroom Help Center, “Recolor an object in a photo,” updated 29 August 2026; Photoroom, Recolor product page, accessed 29 August 2026; Google AI for Developers, Gemini image generation and semantic masking documentation, updated August 2026.