Photo Enhancer
Turn a soft, grainy or blocky photo into a clearer one. The AI cleans up compression, sharpens edges and brings back texture, on your own device. Free up to 1024 px, HD at full resolution.
Drop a photo to enlarge it
or paste with Ctrl+V · JPG PNG WebP HEIC GIF · max 25 MB
Your photo is enlarged on this device and never uploaded. Faces are not retouched: the model sharpens what is there.
How to
How to enhance a photo
- 1
Drop your photo
JPG, PNG, WebP or an iPhone HEIC photo. The free result is made at once, on your device.
- 2
Compare
Drag the slider to see the original and the result side by side.
- 3
Download
Save the free result, or make the HD version at full resolution (same size, 2x or 4x).
Why this size
What the photo enhancer fixes, and what it leaves alone
Most photos look poor for three reasons: they were made small, they were compressed by a chat app or a website, or the camera did not quite focus. The enhancer runs an AI model trained to undo exactly that. It removes JPEG blocks and grain, makes edges crisp again and rebuilds fine texture such as hair, fabric, grass and brick.
It is not a beauty filter. Faces are sharpened like everything else, never smoothed, slimmed or replaced, so the person still looks like themselves. It does not change colours or light either: what you get is the same photo, clearer.
Choose "Same size" in HD to keep the photo's dimensions, or 2x and 4x to make it bigger at the same time.
Use cases
Photos people enhance most
Photos from chats
Pictures sent through messaging apps lose detail to compression. The enhancer cleans the blocks and grain.
Old phone photos
Small, soft photos from older phones come back sharper and larger.
Product photos
Low resolution product shots get crisp edges for a shop listing or a catalogue.
Screenshots and scans
Blurry screenshots and low resolution scans become easier to read and print.
Tips
Tips that make a difference
Use the clearest copy you have
It cannot invent what is not there
FAQ
Photo Enhancer: questions
Checked
How we checked this page
Checked by the SizePort team on .
We compare every size and rule on this page with the official sources below, and test the tool with sample photos before it goes live. Rules and platforms change, so if something here is out of date, tell us and we will correct it.
How the tool works
An open AI model, NERVE (Apache 2.0, trained only on public domain photos), downloads to your browser once (7 MB) and runs there with WebGPU, or on the processor where WebGPU is missing. First the tool measures how sharp your photo is. A sharp photo is given to the model at full detail and only cleaned up, with part of its own pixels kept so it stays natural; a soft or blurry one is given at a smaller size, where the model can rebuild the edges the blur hides. You can also choose Gentle or Strong yourself. The model works at four times the size in overlapping tiles, so any size works without seams, and its result is averaged down to the size you get. The model file is checked against its fingerprint before it runs. Faces are not retouched, and colours and light are left as they are. Everything happens in your browser; your images are never uploaded.