How to enhance a photo or picture — 3 methods that work

1. AI upscaling — let the model redraw the detail

When a picture is too small or too soft, stretching it only spreads the existing pixels over a larger area. An AI super-resolution model like Real-ESRGAN does something different: it was trained on millions of high-resolution images paired with degraded copies, so it learned what sharp edges, textures and fine gradients look like at higher resolution. At inference time it reconstructs that detail instead of interpolating a blur.

In practice this means a 640×480 snapshot can become a clean 2560×1920 image, and a 960×540 web thumbnail lands at exactly 4K (3840×2160) at the 4× setting. The result is not a pixel-perfect recovery of the original scene — it is a plausible reconstruction — but compared with bicubic upscaling the difference on edges and textures is not subtle.

You can try this right now in the tool above or on the enhance picture homepage: drop a photo and the model starts on its own, on your own device — no upload, no signup, no watermark. If your picture is already the right size but looks compressed or noisy, the enhance picture quality page runs the same AI at 1× to clean it up without enlarging it. How far upscaling can go, and where it stops, is covered in the guide to enhancing picture resolution.

2. Auto-correction — fix exposure, colour and contrast

Not every bad picture is low-resolution. Many are the right size but look flat because the camera exposed for the whole frame: the sky ate the contrast, skin went grey, and shadow detail disappeared into a muddy middle tone. These need corrections, not upscaling.

Auto-levels is the fastest fix. It finds the darkest and brightest 0.4% of pixels in each colour channel and stretches the range back to full black and full white. The result usually pops in a way that looks obvious in hindsight — the data was there, it just was not using the full tonal range.

On top of that you can layer a contrast curve, a luma-preserving saturation boost (so skin stays believable), and an unsharp mask for soft edges. The enhance picture tool keeps these under Fine-tune: tick Auto levels for the stretch, then drag Brightness, Contrast, Colour and Sharpness if the automatic choice needs a nudge. They apply on top of the AI result — pick 1× clean up when the size is already right — and the before/after split updates as you drag.

3. Manual adjustment — fine-tune the result yourself

Sometimes the automatic pass gets 90% right and the last 10% needs a human eye. A sunset may want less saturation correction than the algorithm picks; a scanned document looks better with sharpness cranked up higher than the default.

After the AI pass, use the Fine-tune controls to dial in exactly the look you want:

• Sharpness — increases edge contrast via unsharp mask. Good for text, architecture, product photos. Overdo it and halos appear around edges.
• Brightness — shifts the midpoint up or down without clipping highlights or shadows.
• Contrast — widens the gap between light and dark tones. A small lift makes most photos feel more three-dimensional.
• Colour — saturation: intensifies colour. Keep it subtle on portraits (skin goes orange fast) but push harder on landscapes and food.
• Auto levels — toggle the per-channel stretch on or off to compare.

Every adjustment updates the preview in real time and nothing saves until you hit Download, so there is no cost to experimenting.

Which method should you use?

The answer depends on what is wrong with the picture:

• Too small or blurry — AI upscaling at 2× or 4×. Start on the enhance picture resolution page, which opens at 4× and shows the exact pixel dimensions before you download.
• Right size, dull or flat — 1× clean up with Auto levels and a contrast lift.
• Both problems — 2× or 4× handles both: the model upscales and reconstructs detail in one pass, then the sliders refine colour and tone on top.

All three run entirely in your browser. The AI model (4.9 MB, cached after the first load) executes through WebGPU on the graphics card or falls back to WebAssembly, which is also the path on Safari and every iPhone. No pixel ever leaves your device, there is no daily quota and no account to create. Very large photos are scaled down before the AI pass — to 4 megapixels with WebGPU (2.5 MP at 4×), 0.7 MP without it — and the tool shows the size it will read.

Step-by-step: enhance a picture with the tool above

1. Drop your photo on the tool at the top of this page, tap Choose pictures, or paste with Ctrl+V (⌘V on Mac).
2. Pick the size: 1× clean up does not enlarge, 2× larger and 4× larger multiply the dimensions. The AI starts on its own.
3. Drag the before/after split to compare. Tick Auto levels or move the sliders if needed.
4. Press Download enhanced — or Copy to paste directly into another app.

Up to 20 pictures fit in one batch. The one in the preview is enhanced first; the others run when you open them or press Download all.

What enhancement cannot do

AI upscaling adds plausible detail, not the original detail. Text that was too small to read in the source stays unreadable. Small faces in the background may come back looking subtly different. Severe motion blur or heavy JPEG artefacts leave traces even after enhancement. When the source is genuinely destroyed, a cleaner-looking version is the best you can get — not a forensic recovery.

For most everyday photos — thumbnails from messaging apps, old email attachments, web exports, screenshots of something too small, phone pictures taken in poor light — the improvement is immediate and obvious. That is the use case enhancement tools are built for. If you would rather weigh other free tools first, the overview of free AI photo enhancers covers four of them and where each one processes your photo.