Common retouching tasks for clothing photos
| Task | What it fixes | By hand | With AI generation |
|---|---|---|---|
| Color correction | A garment that looks warmer, cooler or darker than it is | Adjust white balance and color against the real item | The AI copies the color it sees, so the source photo must already be accurate |
| Dust, lint and hair | Specks that look like dirt when zoomed | Spot healing, image by image | A newly generated image usually does not carry them over; check close-ups |
| Creases and wrinkles | Folds from packing or hanging | Steam before the shoot, then smooth what is left | The garment is generated on a model or in a scene; keep sharp source photos |
| Background cleanup | Uneven, grey or cluttered backgrounds | Mask the product and replace the background | Choose a scene or switch the model off for a product on a clean background |
| Shape and symmetry | Twisted seams, uneven sleeves | Liquify and warp, carefully | Not a retouching step; generated images follow the garment in your photos |
| Ghost mannequin | Removes the mannequin and joins front and inside views | Composite of two or more shots | Choose the Transparent mannequin model in AI Studio for the worn shape on a see through mannequin |
| Crop and size | Different crops and sizes across products | Crop every image to the master size | Set the quality once per generation; crop if your store needs a fixed shape |
Not every photo needs every task. A well lit studio shot on a clean background may need only a color check and a crop, while a phone photo of a knit sweater on a bed needs most of the list.
What retouching is, and what it is not
Retouching is editing a product photo after the shoot so it shows the product accurately and looks like the rest of your catalog. The goal is a photo that matches what arrives in the parcel, only cleaner.
It is not making the product look different from what it is. Changing the shade of a fabric, slimming the cut of a dress or hiding a real flaw on a pre-owned item crosses into misrepresentation. Marketplaces are clear about this. Google Merchant Center asks that the main image clearly displays the exact item being sold, that the photo accurately reflects its exact color, pattern, finish and material, and that each variant has its own image. It also forbids promotional overlays, watermarks, borders and added text on product images.
So a useful test for every edit is simple: would a shopper who opens the parcel feel the photo was honest? If yes, the edit is retouching. If not, undo it.
A retouching workflow for clothing
Choose the frames
Pick the sharpest shot for each view and discard the rest before you edit anything.
Correct exposure and color
Set white balance from a neutral area, then compare the garment on screen with the real item in daylight.
Clean the background
Even out the background or replace it, keeping a soft natural shadow if your catalog uses one.
Remove dust, lint and stray threads
Zoom in to full size and work across the garment in a grid so nothing is missed.
Smooth creases
Reduce folds that come from packing, but keep the natural drape and the fabric's texture.
Crop and export
Crop to your master size and aspect ratio, export in a web format and name the file.
Do the color step on a calibrated or at least well set up screen, and check the result on a phone as well, since most shoppers will see it there. For size and naming, see our guides to the best image size for an ecommerce website and image file names.
Color accuracy comes first
Of all retouching tasks, color matters most in fashion. A shopper can live with a small crease in a photo, but a burgundy sweater that arrives brick red feels like the wrong product.
A few habits keep color honest:
- Shoot in steady, neutral light. Daylight from a window or a pair of matching lamps is easier to correct than mixed light.
- Keep the real garment next to the screen while you correct, and compare in daylight.
- Edit each color variant from its own photo. Recoloring one photo into several colors rarely matches the real fabrics.
- Use the same correction on every photo of one shoot, so a collection does not drift from warm to cool.
- Avoid heavy filters. Filters that look good on social media shift colors on product photos.
Wizia's help center makes the same point about source photos for AI generation: shifted color and contrast carry over into the generated shots, so a filter on your input means the customer sees the product in the wrong color.
Retouching by hand: in house or outsourced
Many small stores edit in house with a photo editor, which works well when the catalog is small and the shoots are consistent. The costs are time and skill: removing dust and creases from a full collection, image by image, takes hours, and quality depends on who does it.
Retouching services take that work off your hands and usually charge per image, with higher rates for complex jobs such as ghost mannequin composites or detailed jewelry. Prices vary widely by country and by task, so ask several services for a quote on the same three sample images, including one difficult one. Check how they handle color: a service that has never seen the real garment can only match color to a reference you send.
Whichever you choose, write down your rules once: background color, shadow or no shadow, crop and size, how much crease removal is acceptable. Our article on consistent catalog photos covers what such a photo guide should include.
Where AI generation replaces part of the work
AI generation changes the question. Instead of fixing a photo pixel by pixel, you create a new image from clean source photos. In Wizia AI Studio you upload a few photos of a garment from different angles and get studio product images and on-model photos in a scene you choose. You can also switch the model off, and the product is generated on its own on a clean background.
That removes several retouching steps at once: background cleanup, much of the dust and lint, and the consistency work, because all shots of a product come from one session with the same settings. It does not remove the need for good input. Wizia's help center is explicit that the AI copies exactly what it sees, color, cut, fabric and details, and recommends daylight, a plain background, the whole product in frame, no heavy filters and several angles with close-ups.
Wizia does not retouch your original photo file. When a generated frame is not right, you regenerate it and describe only what should change, for example a brighter background or a logo that must be readable, and you can attach an extra photo of a detail that keeps getting lost. The first regeneration of each frame is free, up to 3 per generation, and Pro and higher plans include unlimited regenerations. A paid regeneration costs the same as a new image: 1 credit at 1K or 2K and 2 credits at 4K.
Check every image before it goes live
Whether a person or an AI made the final image, check it against the garment before publishing:
- Color: compare in daylight with the real item.
- Details: zoom into logos, prints, buttons, zips and stitching. They should match the product exactly.
- Shape and length: the hem, sleeves and neckline should look like the real garment, not an idealised version.
- Variants: each color has its own image, attached to the right variant.
- Overlays: no text, badges or watermarks on images meant for marketplaces.
If you sell through Google Shopping, also keep the metadata of images made with generative AI: Google Merchant Center asks that such images contain metadata saying they are AI generated and that you do not remove tags such as IPTC DigitalSourceType. Some compression and editing tools strip metadata by default, so check their export settings. Wizia exports carry no IPTC or C2PA AI metadata today, so ask support@wizia.ai before you send them to Merchant Center. More marketplace rules are in our Google Shopping image requirements guide.