Recolor an image online by sampling the exact shade you want to change. The instant preview runs in your browser for free, and one optional AI pass repaints textured objects while keeping the original lighting, reflections, and material detail.
Source image / 01
Upload a photo, then click the surface whose color you want to change.
JPG, PNG, or WebP · 10 MB max · the free preview never leaves your browser
Result / live preview
Your recolored image will appear here
The first screen is the workbench. Upload a photo, click the shade you want to change, and compare the result instantly.
Recolor studies
These visual studies show the decision this tool is built for: hold the whole photograph still and move a single color, so the buyer sees the option they actually care about.
A colorway test keeps the model, pose, and studio lighting while the garment color changes, so a product page can show the same shot in more than one finish.
ReviewInspect seams, zips, embroidery, and shadow transitions where a machine stitch sits next to the fabric.
Furniture buyers need to see a shade inside their own room. A recolor keeps the room, the window light, and the rug exactly where they were.
ReviewCheck the fabric weave, cushion piping, and the shadow under the frame before the image reaches a customer.
Packaging mockups live or die on an exact brand color. Recoloring the container keeps the label, reflections, and glass highlights readable.
ReviewVerify label text, the glass highlight, and the actual paint or ink standard against the source file.
Definition
To recolor an image is to change the color of a defined part of a photo while leaving everything else in place. A recolor image tool answers three questions separately: which pixels are allowed to change, what new color they become, and how the object keeps its original shading so the edit still looks photographed.
Most disappointing recolors fail on the third question, not the first. A flat fill drops the highlights, contact shadows, and reflections that made the surface believable, so a jacket turns into a sticker and a glass bottle turns into plastic. Keeping each pixel’s own lightness while swapping its hue is what preserves the material.
This page splits the job by cost and precision. Sampling a shade, choosing a tolerance, and previewing the replacement happen locally in your browser and cost nothing, which is enough for flat and mostly-solid surfaces. One optional GPT Image 2.5 pass then handles the harder cases: print, knitwear, brushed metal, glass, and any object whose color varies across the frame.
Practical workflow
The sequence is deliberately short. The important decisions are visible in the workbench, and you can review every change before you spend a credit.
Choose a JPG, PNG, or WebP file up to 10 MB. The browser reads the real pixel dimensions before anything is sent, so a file that only pretends to be a supported image is rejected immediately.
Use a source where the target surface is clearly visible and not hidden behind another object.
Click the eyedropper on the photo to sample the exact pixel you mean, or start from the six shades detected in the image itself. Then pick a target color from the curated swatches or type any hex value.
If two different objects share the sampled shade, raise the region label with a plain description such as “the left cushion”.
The free in-browser pass repaints matching pixels instantly, with a 0–100 tolerance slider and a draggable Before / After review. Textured surfaces can then be refined by a single GPT Image 2.5 task that costs 15 site credits.
Compare the result against the original at full size and confirm the untouched areas really are untouched.
Method selection
Two production passes ship on this page, and one common shortcut deliberately does not. Choose the method that matches the surface in front of you instead of pushing one method harder.
| Method | How it works | Where it wins | Where it fails | Cost on this page |
|---|---|---|---|---|
| Browser replace color | Every pixel within the tolerance of the sampled shade takes the new hue while keeping its own lightness | Flat walls, painted metal, solid fabric, and graphic shapes change immediately and exactly | Gradients, print, and soft shadows may keep a leftover rim of the old color | Free · no credits |
| Whole-image hue shift | A global rotation of every hue in the frame, applied to the entire photo at once | Fast when the real goal is “shift all the colors together”, such as a quick mood change | Unrelated colors drift too and neutral areas turn muddy, which is why this page does not use it for a single-object recolor | Not offered on this page |
| GPT Image 2.5 recolor | One image-to-image pass repaints the described object and rebuilds its shading and texture | Knitwear, glass, brushed metal, reflections, and objects whose shade varies across the frame | A very vague region description can leave neighbouring objects slightly tinted, which is why the sampled hex and a region label are sent together | 15 site credits per 1K result |
No automatic recolor is a guarantee, and this page does not offer a brush mask. Reflection, translucency, and fabric print deserve a human check at the size the image will actually be used, especially when the result carries a product or a brand color.
Use cases
A recolor is a sales and communication tool before it is an editing trick. Each of these jobs changes one color on purpose and protects everything else.
Does a shopper need to see the same product shot in another finish before deciding?
Try this: Sample the garment or product body itself, not a shadow, then compare the recolored version with the original side by side at thumbnail size.
ReviewConfirm that labels, logos, stitching, and packaging text still read exactly as they did in the source.
Would a customer commit faster if they could see the sofa, cabinet, or wall in their own photo?
Try this: Keep the room light untouched and move only the upholstery or paint. A moderate tolerance holds the fabric weave while the surrounding trim stays neutral.
ReviewCheck the shadow under the furniture and any reflection in a window or screen before you publish the rendering.
Does the container need to match a fixed brand standard or a seasonal palette?
Try this: Type the brand hex value instead of eyeballing a swatch, and use the AI pass when the surface is glass, foil, or a textured label.
ReviewCompare the final hex against the brand standard and verify that the highlight and reflection shapes were not redrawn.
Commercial comparison
Each of these products can recolor a photo, and they are optimized for different moments. The comparison below focuses on the decision this page makes: how fast you can go from a sampled shade to a reviewable image without installing anything. Competitor names are named for clarity; this page does not link out to them.
Workflow comparison reviewed
| Decision point | Nano Banana Pro | Photoshop | Canva |
|---|---|---|---|
| Time to first result | Upload, click the surface to sample its shade, and the recolored preview appears immediately in the browser. | Requires an installed desktop application, a layered document, and a selection or adjustment layer workflow. | Runs in a browser inside a design project, so a quick one-off edit competes with layout tools you did not need. |
| Precision when the color varies | Send the sampled hex, a region label, and a scope choice to one GPT Image 2.5 pass that rebuilds shading and texture. | Very deep, provided you know which masking and blend approach matches the surface in front of you. | Convenient for simple flat shapes; complex reflective or printed surfaces usually need a different tool. |
| Untouched areas | The instant pass only writes to pixels inside the tolerance band, and the AI pass is instructed to change nothing outside the described area. | Whatever stays outside your selection stays intact, and the selection is entirely your responsibility. | Depends on the selected element and the tool you pick, so results vary between simple and complex photos. |
| Cost of experimenting | Sampling, tolerance, swatches, the live preview, and PNG export are free; only the optional AI refine costs credits. | Sold as a subscription; the trial period and license terms depend on your current plan. | Free tier plus paid plans, where feature and export rules depend on the current plan and account. |
| Best fit overall | A single, reviewable color change on a web page, from a phone or a laptop, without an editing suite. | Professional retouching where color is one step inside a larger, tightly controlled pipeline. | Design production that already lives inside a template, a deck, or a campaign layout. |
Choose Nano Banana Pro when the job is one honest color change and you want to see it before you pay for it. Choose a desktop editor when you need pixel-level control across many layers, or a design suite when the recolor is one step inside a bigger composition.
Operating facts
These are the boundaries the public workbench actually enforces, not marketing promises about the result of any single edit.
The instant pass is free. One optional AI refine costs 15 site credits and the region label is capped at 60 characters.
The browser decodes your file locally to measure it and to build the sampled palette. Choosing a shade, moving the tolerance, switching swatches, and exporting the instant PNG all happen on your device and never leave the page.
The AI refine is the only step that uploads anything. The server checks the real image container, size, dimensions, and moderation result, stores the file under your signed-in account, and verifies that the stored reference belongs to you before the task starts. Task state, credits, and refund handling follow the existing workbench flow.
Credits are reserved and consumed by the server for the AI task only. Sampling colors, changing the target swatch, adjusting tolerance, and downloading the instant PNG do not create a task.
Common questions
These answers separate what the page implements from editorial advice about choosing and reviewing a color change.
Yes. Uploading, sampling the shade, choosing a target color, adjusting tolerance, viewing the Before / After preview, and downloading the instant PNG all run in your browser and cost nothing. Only the optional GPT Image 2.5 refine starts a paid task.
Yes. The free local pass is a deterministic pixel replacement: every pixel inside the 0–100 tolerance band takes the new hue while keeping its own lightness. Flat paint, solid fabric, and graphic shapes usually need nothing else.
Click the eyedropper on the surface you mean and the exact pixel color is sampled. The tolerance slider then decides how far a neighbouring pixel may drift and still be included, and an optional region label such as “the left cushion” is added to the AI instruction for extra clarity.
The local pass keeps each pixel’s own lightness, so folds, gradients, and contact shadows survive. The AI pass is explicitly instructed to preserve shading, highlights, reflections, and material finish, and to change nothing outside the described area.
A color that sits just outside the tolerance band is left alone, which is common on soft shadows, gradients, and anti-aliased edges. Raise the tolerance for a wider match, lower it when two objects share a similar shade, or use the AI refine for surfaces whose color varies across the frame.
One AI refine costs 15 site credits at 1K and returns a downloadable image file. The instant export is a real PNG encoded from the canvas pixels rather than a screenshot of the interface, and it is written at the loaded image resolution instead of a fixed preview size.
Return to the color lab
Start with a photo you can inspect, sample the shade you actually mean, and preview the swap for free. Reach for the AI pass when the surface is textured, reflective, or printed, and review the result before it ships.