Image Sharpener
Crisp up soft edges. Pick a style, choose a model, and download the result at full resolution with no watermark.
Pick a style on the left, then hit Apply. Your original stays untouched — you always see both.
How to use Image Sharpener
Scan or photograph the print
A flatbed scan at 600 dpi is ideal. A phone photo works: lay the print flat, use indirect daylight, avoid your own shadow and any glare from the surface.
Choose the repair
Repair damage, correct faded colour, colourise a black-and-white original, or describe what the photo needs in a sentence.
Compare and download
Check the result against the scan at full size, then download at full resolution. Keep both — the scan is the document, the restoration is the reading of it.
See it in action — real before & after
Why use Image Sharpener
What Image Sharpener does
Crisp up soft edges.
Faces treated with care
The face is the reason the photo is being restored. Features are reconstructed conservatively rather than replaced with a generic plausible face.
Grain and noise handled properly
Film grain is reduced without smearing away the fine detail that sits at the same scale — the usual failure of automatic denoising.
Colourisation available
Black-and-white originals can be colourised with period-plausible tones, and the monochrome version is still there afterwards.
Print-ready resolution
Enough resolution to reprint and frame, which is what most people restoring a photo intend to do with it.
The original is never altered
Your scan stays on your device untouched. The restoration is a separate file.
What Image Sharpener changes
Sharpen this image: increase acuity on genuine edges and detail, leave flat areas free of amplified noise, and avoid visible halos around high-contrast boundaries.
Your face, bone structure and skin tone are held fixed on every generation — only what is described above is changed. Add your own brief in the Custom tab to take it further.
Image Sharpener for the damage prints actually have
Different decades fail differently. A 1940s silver print, a 1975 colour snapshot and a 1998 disposable-camera photo each need a different repair.
Damaged group and wedding photos
Often the only surviving image of several people, which makes conservative face reconstruction the priority.
Torn and creased prints
Fold lines, corner losses and surface scratches from decades in an envelope or an album.
Faded colour from the 60s to the 80s
Dye layers fade at different rates, which is why these photos go magenta. Reversing the shift is the single biggest improvement available.
Black-and-white originals
Contrast restored, dust and scratches removed, and colourisation offered as an option rather than forced on you.
Key highlights of Image Sharpener
Benefits
Colour and monochrome versions
Colourise for impact, keep the monochrome for accuracy. There is no need to choose one.
No subscription
A one-time credit pack, credits that never expire, and no renewal.
Nothing to install
It runs in the browser on a phone or a laptop. No desktop app, no plugin, no account needed to look around.
Two models to choose from
Nano Banana and GPT Image handle the same job differently. If one result is not right, the other model often is.
Every aspect ratio
Square, 4:5, 9:16, 16:9 and more, chosen before generating so nothing important gets cropped away afterwards.
What people use it for
What it is mainly used for
Crisp up soft edges. Pick a style, choose a model, and download the result at full resolution with no watermark.
Anniversary and birthday gifts
A restored and reprinted wedding photo is a better gift than almost anything bought new.
Genealogy and family archives
Restored images attached to a family tree, alongside the unaltered scans.
Local history and archive projects
Faded documentary photographs made legible for display or publication.
Damaged wedding albums
Albums from the 70s and 80s where the adhesive pages have discoloured the prints themselves.
Prompt ideas
Describe the damage you can see and say what must not change. Restoration prompts fail when they are ambitious — asking for a photo to be "improved" invites the model to make a new one.
Ask for era-plausible colour
"1950s colour palette, muted, no modern saturation" avoids a restoration that looks like a photo taken last year.
Address grain deliberately
"Reduce grain but keep detail in the fabric and hair" gets a better result than asking for it to be removed.
Mention the background separately
"Repair the water stain in the upper right of the background" isolates a fix that would otherwise affect the whole frame.
One area at a time
Repair the worst damage first, look at the result, then address the next. Compound requests compound the errors.
Be concrete, not complimentary
"Soft light from the left, plain grey background" gives the model something to act on. "Beautiful, high quality, 8k" gives it nothing at all.
Change one thing at a time
If a result is close, edit a single detail and run it again. Rewriting the whole prompt gives you an unrelated image and no information about what the change did.
More about Image Sharpener
Photographic prints are chemistry, and chemistry has a shelf life. Dyes shift, silver tarnishes, gelatine cracks, and damp does the rest. Most family boxes hold photographs that are twenty years from being unreadable. Image Sharpener is for the ones still legible enough to work from.
There is an important distinction between restoration and improvement, and it is worth holding onto. Restoration returns a photograph to what it looked like when it was made; improvement makes a nicer picture. Where damage destroyed information entirely, what fills the gap is inference, and it should be recognised as such.
The practical bottleneck is almost never the model — it is the scan. A 600 dpi flatbed scan of a print, taken out from behind glass, gives a result that a phone photo shot at an angle in a lit room cannot approach, however good the restoration is.
Getting the best result from Image Sharpener
- Judge every result at full size, not zoomed out. Problems hide at small sizes.
- Generate two or three times before deciding. Results vary between runs, and the first is not reliably the best.
- Check hands, text, jewellery and glasses in any result before sharing it — those are still where image models most often slip.
- Choose the aspect ratio before generating rather than cropping afterwards.
- If one model ignores part of your instruction, switch to the other and run it again.
- Keep the original file. It is a better base for the next attempt than a generated image is.
- For a matching set, keep the same source photo and style and change only one variable.
- Nothing you upload is published, shared or added to a gallery — but do not upload anything you would not want to leave your device at all.
It's free to try and runs online — nothing to install. Explore the rest of our AI Creative Apps, or open the full AI photo editor for more control.
