The Img ai Image Colorizer reads a grayscale photograph and predicts the colors that plausibly fit what the camera recorded. Every hue is inferred rather than recovered, so the original stays beside the result for comparison.
Color inference
Upload one grayscale image, run a single automatic pass, then compare the inferred color against the original before you keep it.
Automatic colorization
One inference pass over the whole frame. There is no palette picker, strength slider, or region brush, because this backend does not expose them.
Source
—
Output
1K render · same framing
Your colorized photo will appear here
Upload a grayscale JPG, PNG, or WebP between 256 and 4,096 pixels per side to check it and prepare a colorization pass.
Concept examples, not measured customer results.
The direct answer
An image colorizer adds color to a black-and-white or grayscale photograph. It does not read a hidden color layer, because none exists: a grayscale pixel stores brightness only, and many different colors share the same brightness. The model instead recognizes what a region probably is — a wool coat, a brick wall, summer leaves, an overcast sky — and assigns color that is consistent with that reading and with the rest of the frame.
That is why a colorizer can be convincing and wrong at the same time. A red dress and a mid-green dress can be identical in grayscale, so the result reflects statistical likelihood, not evidence. Img ai treats the output as a visual interpretation you review, which is why the workbench keeps the grayscale source in the same view and why this page never claims historical color accuracy.
Scanned portraits are the most common colorization job and the most demanding one. Faces carry the strongest expectations: skin, lips, eyes, and hair have to sit in a believable range or the whole frame reads as fake. Use the split view on the face first. If a relative wore a specific uniform, habit, or wedding dress, treat the generated color as a placeholder and check it against family knowledge before printing or sharing it.
Archival street photography compresses shopfronts, signage, vehicles, and crowds into one grey field. Adding color separates those planes and makes the scene easier to read at a glance for a blog post, a local-history talk, or a slide. Signage color and paint schemes are exactly where inference is weakest, so caption the image as colorized and keep the original file alongside it.
Photographers also shoot in black and white on purpose, and sometimes need to see what a scene would look like in color before deciding how to grade or reshoot it. A colorized pass is a fast reference for that decision. It is not a color-managed conversion, it will not match a raw file, and it should not be used as a proof for print work.
Sign in, then add a single JPG, PNG, or WebP up to 10 MB with each side between 256 and 4,096 pixels and no more than 16 megapixels. Img ai checks the file on the server as well as in the browser, so a mislabelled or truncated image is rejected before anything is charged. Crop out album borders and heavy dust first; the model treats damage as part of the picture.
The workbench shows the fixed 10-credit cost and your remaining balance before submission. There is no palette picker, strength slider, or region brush on this page, because this backend does not expose them and a control that does nothing would be a lie. One automatic colorization over the whole frame is exactly what you are buying.
Drag the divider across the result and inspect faces, hands, clothing seams, foliage, and any painted surface. Download when the interpretation is good enough for how you plan to use it. If the connection drops or you refresh the tab, Img ai recovers the same task identity instead of starting a second paid job, and the run stays visible in My Creations.
Concept examples
These owned concept pairs show where an inferred palette is usually safe and where it is a guess: a studio portrait, a busy street, and an open landscape. They are not measured Img ai customer results.
Portraits are judged on skin first. Check that tone stays even across the cheek and jaw, that lips are not pushed toward makeup that was never there, and that hair keeps a single believable shade. The garment color here is a guess, and a plausible guess is still a guess.
Automatic pass · 1K render · 10 credits
Concept example generated for this page; not a measured customer result from Img ai.
Street scenes show what colorization is good at and where it drifts. Brick, asphalt, foliage, and sky usually land in a sensible range. Awnings, vehicle paint, and any lettering are invented, so never present them as a record of how the street actually looked.
Automatic pass · framing preserved · no signage rewriting
Concept example generated for this page; not a measured customer result from Img ai.
Landscapes are the friendliest input: water, vegetation, and rock have narrow plausible ranges, so the pass tends to look natural. Watch the horizon and the waterline for color bleeding between sky and sea, which is the most common landscape artifact.
Automatic pass · grain and scratches left untouched
Concept example generated for this page; not a measured customer result from Img ai.
The page exposes only the automatic operation this backend supports. A palette picker, intensity slider, and per-region brush are absent because they are not independent controls in this workflow.
One pass assigns color across the entire image, keeping subjects, edges, crop, and lighting pattern as photographed. It is designed to look like a photograph rather than a tinted illustration.
Skin, hair, textiles, foliage, water, stone, wood, and metal usually land in a coherent range for one another, which is what makes a scene read correctly at a glance.
The grayscale source and the colorized result sit in one comparison view, and the finished task stays in My Creations so you can return to it instead of re-running a paid job.
Three different problems get confused with each other constantly. Pick by what is actually wrong with the file in front of you.
| The problem | Image Colorizer | Photo Enhancer | A prompt-driven editor |
|---|---|---|---|
| Best for | A clean grayscale photo that needs color | A color photo that looks soft or compressed | A specific object or area you want changed |
| What changes | Color only; structure is preserved | Perceived clarity; size and color stay | Whatever the prompt and selection cover |
| What it will not do | Repair damage or prove historical color | Add color to a monochrome source | Guarantee an untouched background |
| Cost on Img ai | 10 credits | 11 credits | Depends on the tool |
Clean scans, well-exposed monochrome negatives, and modern black-and-white captures give the pass the most structure to read. Even lighting, visible material texture, and a frame that has not been crushed to pure black or pure white all improve the result.
These three tools answer the same query in different ways: a tracked single-pass colorizer inside a wider image account, a filter-driven creative colorizer, and a colorization-plus-restoration specialist. Public product pages were reviewed on 6 September 2026; confirm current plan terms on each vendor site before you commit.
One tracked colorization pass inside the same account, credits, and history you already use for the rest of Img ai.
Sign in, upload one JPG, PNG, or WebP up to 10 MB, confirm 10 credits, run a single automatic pass, drag the split comparison, and download. The run is recoverable and appears in My Creations.
No palette filters, no reference-image steering, no batch queue, and a 1K-class render rather than a large-format export.
Open the Img ai toolCreative direction over the mood of the color rather than a single neutral interpretation.
Palette.fm advertises "21+ Color filters" plus keyword customization, unlimited previews capped at 500x500 with a watermark, and one free full-resolution credit; paid output goes up to 5000x5000.
The free tier is a watermarked low-resolution preview, and credits are bought as a separate subscription or pack outside your existing image workflow.
Public product pageColorization bundled with old-photo restoration when the scan is damaged as well as monochrome.
ImageColorizer offers V0, V1, and V2 models and states that you can "customize the positive and negative prompts and control the color palette", with a free account plus credits and limits that "may vary by plan".
More knobs mean more decisions and more runs, and the upload, dimension, and batch limits are plan-dependent rather than a single published number.
Public product pageAlso worth knowing: MyHeritage describes In Color as the "world's best technology for colorizing and restoring the colors in historical photos", and it is the natural choice if your photos already live in a genealogy tree there. Img ai is the better fit when colorization is one step in general image work rather than a family-history project.
Upload only photos you own or are authorized to process, and be careful with images of identifiable people who cannot consent. Img ai keeps the tracked task on your account and the result in My Creations. Read the privacy policy before uploading family or archival material.
No. Grayscale brightness cannot identify a hue, so several colors fit the same pixel equally well. The result is a plausible interpretation. When accuracy matters, check uniforms, packaging, vehicles, and clothing against written records, catalogues, or family memory, and label the image as colorized.
It will run, but this page only adds color. Scratches, tears, dust, and fading remain, and severe damage can push the color reading off. Clean or crop the worst areas first, and treat colorization as a separate step from repair rather than a rescue for a badly degraded print.
One JPG, PNG, or WebP file up to 10 MB, with each side between 256 and 4,096 pixels and no more than 16 megapixels in total. The file is validated in the browser and again on the server, so a mismatched or incomplete image is rejected before any credits are reserved.
The framing and composition are preserved, but the output is re-rendered at a 1K-class size rather than returned at the exact pixel dimensions of a large scan. Keep your original file if you need the full-resolution monochrome version for archiving or printing.
Each run costs 10 credits, shown next to your balance before you submit. A failed task refunds automatically, and a dropped connection or page refresh recovers the same task instead of charging twice. Current plans are listed on the Img ai pricing page.
Not on this page. Img ai exposes only the automatic pass this backend actually supports, so there is no palette picker, intensity slider, or region brush here. If you want directed color changes, take the result into an editor that supports prompt-based or manual adjustments.
First-party limits come from the Img ai colorization backend. Competitor behavior was read from the public product pages below on 6 September 2026, and claims are limited to what those pages state themselves.
Move to the workflow that matches what the source actually needs.
