Which kind of dark is it
Brightening multiplies the tonal values already in the file. It never adds information. So the useful question is not how dark the photo looks, but whether the dark parts still contain differences a curve can pull apart.
Three causes
- Evenly underexposed The whole frame is dim but you can still make out shapes in the shadows. This is the easy case and a restrained lift handles it.
- Backlit subject A person in front of a window, a sky, or a screen. The background is correctly exposed and the subject is a silhouette. Lifting the whole image makes the background white before the face is visible.
- Crushed to black The darkest regions are one flat value. The sensor clipped, or a heavy compression pass flattened them. Nothing there survives, and brightening produces grey noise rather than detail.
A quick test for the third case: zoom in on the dark region on a bright screen. If you can see any variation at all, a curve can work with it. If it reads as one uniform block, it is gone.
Matching the cause to the curve
| What the photo looks like | Profile | What it does |
|---|---|---|
| Dim overall, shadows still readable | Natural | A restrained lift across the whole tonal range |
| Deep shadows hiding real detail | Shadow lift | Opens the darkest regions harder while holding midtones back |
| Dark subject, bright window or sky behind | Backlit | Lifts the subject with the strongest protection on the bright background |
The brightener makes an opening recommendation from a luminance histogram, then shows all three as real previews. The recommendation is arithmetic on tone distribution. It does not identify faces, objects, or the scene, and there is no generative model involved.
The sequence
- Choose the file One still JPEG, PNG, or WebP, up to 20 MB, 8,192 pixels per side, and 16,000,000 pixels total. It stays in the browser.
- Compare the previews Look at the brightest area, not the darkest. That is where a too-strong curve does its damage.
- Process the full image The preview is reduced for speed. The full file is processed once you commit to a look.
- Check before you download The finished file is decoded again and its structure, pixels, luminance, and metadata are checked. Download stays locked until those pass.
Protecting the bright end
Every brightening curve pushes values upward. Values already near white have nowhere to go, so they flatten into pure white and the texture there disappears. Once that happens, no later edit brings it back.
The curves here taper as they approach white for that reason, and a second decode measures how much clipping actually occurred. The download is blocked if the output exceeds the selected profile's limit. That is a guardrail, not a judgement about your photo, so look at the sky or the window in the result before you use it anywhere.
What it costs the file
Brightening changes pixel values, so the output is never bit-identical to the input. A PNG is re-encoded without additional loss. A JPEG or WebP output adds a round of encoder loss on top of the tone change. Keep the original file. If you decide later that Shadow lift was too strong, you want to start from the untouched image rather than from an edit of an edit.
The new copy also drops what was attached: no supported EXIF, GPS, XMP, IPTC, origin, or C2PA records are carried across, and the output is scanned again. No browser scanner can rule out every proprietary or pixel-embedded signal, which is worth remembering if the source was generated rather than photographed. To see what a file carries before you edit it, the metadata inspector lists it, and what EXIF data reveals covers why that matters.
Brightening FAQ
How do I brighten a dark photo?
Work out why it is dark first. An evenly dim photo wants a restrained lift, deep shadows want a stronger shadow curve, and a subject against a bright window wants a curve that protects the background.
Can a completely black area be recovered?
Only if the file still holds tonal differences there. Brightening multiplies what was recorded. Where the sensor clipped or the encoder crushed the values flat, lifting produces grey noise.
Will it blow out the bright parts?
The curves taper near white and a second decode measures the clipping. Download is blocked when the output exceeds the profile's limit, but check the result yourself before using it.
Does brightening reduce quality?
It changes pixel values, so the output is not identical to the original. PNG re-encodes with no extra loss. JPEG or WebP adds encoder loss on top. Keep the original for later edits.
Does this use AI?
No. Fixed tone curves and a luminance histogram. No generative model, no scene recognition, and no invented detail.
Does it work on a phone?
Yes, in a mobile browser with Web Workers, createImageBitmap, and OffscreenCanvas encoding. Large images use more memory on a phone, so the 20 MB and 16,000,000 pixel limits still apply.