Low Light Photo Enhancer

Dragging the exposure slider up in a normal editor also drags up the noise, and the colours go grey. This does the job the way a camera would have if it had had more light: shadows come back, colours come back, and the grain that lives in them is cleaned instead of amplified. Run it free to preview the result; download the clean full-resolution image with credits.

What is wrong with the photo?

Not sure? Leave it on Too dark — it brightens and corrects colour in one pass. Pick Night shot when the photo is both dark and grainy.

Drag & drop a file here, or

JPG PNG WebP Max 20MB • Processed on server • Deleted immediately

Brightness Without the Grain

Raising exposure in an editor multiplies whatever is in the shadows — including the noise. The model was trained on pairs of dark and correctly-lit photos, so it reconstructs what should be there instead of scaling up what is.

Built for Phone Night Shots

Concerts, restaurants, streets after dark, indoor parties — the shots a phone sensor cannot expose properly. Colour casts from mixed lighting get corrected along the way.

Private & Secure

Your photo is sent to our AI server for processing. The enhanced result is held only briefly so you can unlock the HD download, then expires automatically. Nothing is shared.

How it works

Why brightening a dark photo usually makes it worse

A dark photo is not missing brightness — it is missing light, and those are different problems. When too few photons reach the sensor, what the file actually holds in the shadows is mostly sensor noise with a faint signal buried in it. Pull the exposure slider up and you scale both: the picture gets lighter, the grain gets louder, the colours go flat and grey, and skin turns muddy. That is the ceiling of a curves adjustment, and no amount of sliding gets past it. A low-light model works the other way round. It has seen thousands of pairs of the same scene shot badly and shot well, so instead of amplifying the pixels you have, it reconstructs the ones you should have had — separating what was noise from what was detail, restoring the colour the darkness swallowed, and leaving the parts of the frame that were already correctly lit alone. If the photo is properly exposed and simply soft or shaken, that is a different fix — use the Photo Unblur. And if you just want to see the exposure numbers the camera actually used, the EXIF Checker reads ISO and shutter speed straight out of the file.

Getting the best result out of a night photo

Feed it the original file, not a screenshot or a messaging-app copy — compression destroys exactly the faint shadow detail the model needs to work with, and a photo that has already been brightened once has nothing left to recover. Pick the mode that matches the actual fault: Too dark for an underexposed indoor or evening shot, Night shot when it is both dark and grainy, Grainy when the exposure is fine and only the noise is the problem. Expect limits: a frame that is almost entirely black holds no signal to restore, and blown-out highlights — a street lamp, a phone screen, a stage light — are already clipped and stay clipped. Broad shapes and surfaces recover well; tiny text in the dark often does not. Afterwards, run the result through the Image Upscaler if you need it larger, or the Quality Analyzer to check the file is good enough to print. Every run here is free to preview; you only spend credits on the clean, full-resolution download, and the pricing page shows what a credit costs.

Frequently Asked Questions

How do I brighten a dark photo online?

Upload the photo, pick what is wrong with it — too dark, a grainy night shot, or noisy — and the AI rebuilds the exposure: shadows come back, colours are corrected, and the noise hiding in the dark areas is cleaned rather than amplified. The result is free to preview as a watermarked image; the clean full-resolution download costs credits.

Why does making a photo brighter in an editor look bad?

Because the shadows of an underexposed photo contain mostly sensor noise. An exposure slider multiplies every pixel equally, so the noise gets amplified along with the picture, colours wash out and skin looks muddy. A low-light model is trained on pairs of badly-lit and correctly-lit photos, so it reconstructs the missing detail instead of scaling up what is already there.

Is the low light photo enhancer free?

Running it and seeing the result is free, with no account. The preview is watermarked and reduced in size. Downloading the clean, full-resolution image costs credits, which is what pays for the AI processing.

Can it fix a really dark night photo from my phone?

Usually, yes — night shots from a phone are the main use case. What it cannot do is invent light that never hit the sensor: a frame that is almost entirely black has no signal left to recover, and blown-out highlights like street lamps or stage lights are already clipped and stay clipped.

Does it remove noise as well as brighten?

Yes. Low light and noise are the same problem, so the model handles both: the Night shot mode brightens and denoises together, and the Grainy mode denoises only, for photos that are exposed correctly but shot at a high ISO.

Will it change the parts of the photo that are already fine?

It works on the whole frame, but the correction is concentrated where light is missing, so the correctly-exposed parts come back close to how they looked. Colours can shift slightly, since the model also corrects the colour cast that low light introduces.

What AI model does it use?

InstructIR, an all-in-one image restoration model from the University of Wuerzburg and Sony PlayStation FTG, published at ECCV 2024. It is MIT-licensed, so its output is safe to use commercially.

Are my photos stored on the server?

Your photo is sent to our AI server for processing and is not kept. The enhanced result is cached only briefly so you can unlock the HD download, then it expires automatically. Nothing is shared or reused.

Scanly is an independent product. InstructIR is a third-party model we call through Replicate — an integration, not our brand. Scanly is not affiliated with, endorsed by or sponsored by Replicate or the model authors, and all model names and trademarks remain the property of their respective owners. AI Disclosure.