AI Photo Color Correction

A photo can be the right picture and the wrong colour at the same time: a hall under old lighting that went amber, a bathroom that came out green, a backlit face that came out mud. This pass is asked to take the whole frame in hand — white balance to neutral, underexposure and overexposure corrected, saturation and contrast lifted subtly toward a vibrant, natural result.

The colour work arrives with a clean-up attached

Read the built-in instruction and you will see it does more than rebalance. The same request asks for an extreme sharpening pass, for every trace of blur, softness, and haze to be removed, and for noise and grain to be purified out of the surface. That is worth knowing before you point it at three things. A deliberately grainy frame will lose the grain. A moody, underexposed interior will be lifted, because lifting is what it was told to do. And skin, which is where denoising and sharpening disagree most, can come back smoothed while the rest of the frame gains bite. If you want any of that preserved, name it in the prompt box.

Describe the look you are correcting toward

The prompt field asks for the desired subject, style, composition, and important details, and a stated destination beats a list of complaints. "Neutral whites, cool clean daylight, skin left warm" gives the pass something to aim at; "remove the yellow" leaves the endpoint to be guessed. Holding a clear reference in mind helps for the same reason — describe the part of it you want matched rather than the part of your own frame you dislike.

Check the tones in this order

• Skin first. It is the fastest tell of an over-correction, and it is the tone most people recognise as wrong.
• Anything that should read white or grey — a wall, a shirt, paper — which is your neutral anchor.
• Highlights, since a saturation and contrast boost can push them to blocking.
• Shadows, for colour shifts and flattening introduced by noise removal.
• Both edges of the frame. A cast that varies across one image, daylight at one side and a warm lamp at the other, is a mixed-lighting problem, and one global correction cannot serve it.

Frequently asked questions

about colour correcting a photo

Can it fix only half a frame?

The request is written for the image as a whole, so a correction that suits the sunlit side of a room will not suit the lamp-lit side. Where lighting is mixed, check left and right edges separately and expect to describe which region matters more in the prompt.

Why did my dark photo come out lifted?

Correcting underexposure is part of the built-in instruction. Low-key frames lose their shadows as a result. If the darkness is intentional, say so — "keep the background deep, correct only the face" — rather than accepting a brightened version of a picture you wanted quiet.

Does it sharpen as well as recolour?

Yes, and that is deliberate. The instruction bundles a strong clarity and denoise pass with the white balance work, so detail and grain change along with the tones. Compare edges against your upload at full size to be sure the added bite is welcome.

Is one pass enough for a heavily cast image?

Run it once, judge it, then decide. Strong casts can leave a residual tint in the shadows even when the midtones read neutral, and repeated global corrections drift the colours further than a single careful pass. A short, specific prompt usually moves the result more than a second identical run.