AI Teeth Whitening

The instruction here is mostly restrictions. Brighten the teeth so they read cleaner and more radiant, keep the result natural and healthy, make the whitening even and integrated with the light already falling on the mouth and face, avoid over-whitening, and leave the gums and lips unaffected. That is a narrow, localised edit, and its failure modes are just as narrow.

A constrained edit, not a face rework

Because the ask names what must not change — gums, lips, and the existing lighting — a good result is only noticeable when you look at the mouth. If the whole photograph comes back brighter, softer or shifted in colour, the instruction has been read loosely, and the optional Prompt box is where you push back. Its hint invites the details that matter to you, so write something you can check: “match the tooth brightness to the whites of the eyes”, “no pure white, keep the natural variation between teeth”.

Matching the brightness to the light already in the photo

Warm indoor light, golden hour and a direct flash each lay a different cast over a smile. Whitening that ignores that cast produces the two classic tells: blue-white teeth inside an amber photograph, or a brightened smile with a shadow still lying across it. Compare the mouth against other naturally pale things in the frame — a shirt collar, a wall, the eyes — and if those keep their own colour while the teeth jump out of the picture, the pass was too strong.

How close to look at the result

Zoom in and inspect the gum line for a pale halo, the outer tooth edges for brightness bleeding past the lips, the bite line where the upper and lower teeth meet, and the tongue for a colour shift it should not have received. Then view the portrait at the size people will actually see it. At small scale a conservative whitening looks natural and an aggressive one looks false, which is exactly the over-whitening the built-in instruction warns against.

Frequently asked questions

about AI teeth whitening

Is one pass enough?

Run it once and then look. Repeating the same request on its own output compounds, and it drives the teeth toward flat, featureless white, the specific outcome the instruction tells the model to avoid. If the first result is too pale, go back to the original photo and rerun with a softer ask rather than correcting the correction.

Can I ask for a stronger or a subtler result?

The prompt box accepts either, and shade language works better than glow language. “Slightly brighter, keep the translucency at the edges of the front teeth” keeps the surface reading as a real object, while “very white” tends to produce the false look the page was written to prevent. Small wording differences are worth testing on the same source file.

Will this fix other things about my smile?

No. Alignment, chips, gaps and gum shape are outside the request, and the ask explicitly protects the gums and lips. If a crooked tooth is the thing you dislike, a whitening pass will not straighten it, and a brighter surface can draw more attention to the shape underneath.

Why do the edges of my teeth look soft?

Whitening has to find the boundary between tooth and lip, and dark lipstick, a heavy shadow under the upper lip or a low-resolution source all make that edge hard to read. A photo with a wide smile, the teeth large in the frame, and light even across the face gives the best chance of a crisp result.