Enlarge a small or blurry photo and get sharp, clean detail back, for print, signage or the web.
Runs Real-ESRGAN, and GFPGAN for faces, on your own device. Your picture is never uploaded or stored.
or drop a file, or paste an image with Ctrl+V
By uploading an image you agree to our Terms. Your picture is upscaled right in your browser and is never stored.
Each original is a small 384 x 256 px picture. Drag the line to compare a normal enlargement with the AI's 4x.
An AI upscaler makes a picture larger and sharper at once: it adds pixels, then redraws the edges and texture a small or blurry photo has lost. Faces and small text get their own care, so the result looks like the same picture, only clearer.
BeforeAfterProfile pictures, family photos and team headshots usually blur in the face first. Every face in the picture is found and redrawn by GFPGAN, a model trained only on faces, so eyes, skin and hair come back sharp. If the redraw stops looking like the person, the plain upscale is kept instead.
BeforeAfterPhoto upscalers tend to turn tiny letters into made-up squiggles. This one finds the lines of text first and enlarges the small ones smoothly and sharpened, so prices, ingredients and dates keep their real letters. Large headings are redrawn by the AI like the rest of the photo.
BeforeAfterScans of family prints, photos from an early phone camera and pictures saved from WhatsApp are all small and soft. Enlarge them 2x or 4x for a frame, a slideshow or a big screen, with the faces redrawn where they are clear enough. Tears, scratches and faded colour stay as they are.
BeforeAfterMarketplaces want large images: Amazon turns on zoom only from 1000 px on the longer side. Upscale a supplier's small photo and the weave, print and stitching hold up when a buyer zooms in, ready for your listing, Shopify store or catalogue.
A picture saved from WhatsApp or an old website turns into blocks when you stretch it across a menu board. The AI redraws the edges and texture instead, so it stays crisp on a 4K screen or a printed poster.
Each close-up here is the same spot in the same picture: on the left, the original's pixels; on the right, what the HD model made of them.
Drop, paste or choose a JPEG, PNG or WebP up to 10 MB, on a computer or a phone.
Fast works on any device. HD uses the larger model for the finest detail on photos. Faces and small text are handled for you.
Drag the handle, zoom in to check the detail, then save the sharper picture as a PNG or JPG.
A 4K menu screen, a marketplace listing that asks for 1000 px, a festival banner. Enlarge the picture once and it holds up everywhere.
Upload the picture above, choose 2x or 4x, and download the result. The AI redraws the edges and texture the blur has softened, and it is free with no sign-up and no watermark. It works best on pictures that are small, soft or slightly out of focus; heavy motion blur can only be partly undone.
Yes. There is no sign-up, no watermark and no daily limit, because the picture is enlarged on your own device rather than on a paid server.
Upscaling makes a picture larger by adding pixels; unblurring makes it sharper. Stretching a picture the ordinary way only adds blur, so this tool does both at once: the AI fills the new pixels with redrawn edges and texture, and a picture that is blurrier than its size suggests is first reduced to its real sharpness.
Yes. Blocky pixels from a small picture or heavy JPEG compression, such as a photo saved from WhatsApp or a website, are redrawn as smooth edges and real texture.
Yes. A picture 1000 px wide becomes 4000 px wide at 4x, more than 4K's 3840 px. Results go up to 16 megapixels, for example 4096 x 4096.
No, by design. Faces are redrawn by a model trained on faces and then compared with the original; when the redraw no longer looks like the person, the plain upscale is kept. Faces that are already sharp are left alone.
It makes old and scanned photos larger and sharper and redraws faces that are clear enough. It does not repair tears or scratches or bring back faded colour.
Yes, in the browser on Android and iPhone, with no app to install. Fast works on any phone; HD and the face fix need a browser with WebGPU and run best on a computer.
Real-ESRGAN, the open model behind many professional upscalers. Fast uses its compact general model (5 MB); HD uses RealESRGAN x4plus (64 MB), which brings back more texture on photos but takes longer.
A blurry face cannot be rebuilt from its own pixels. So every picture is checked for faces (YuNet), and each small or blurry one is redrawn by GFPGAN v1.4, a model trained on faces, then blended back into the upscaled picture. Faces that are already sharp are left as they are, and so is a face too turned or blurred to redraw faithfully: when the redraw no longer matches the real face, the plain upscale is kept. It needs a browser with WebGPU and downloads a 170 MB model the first time a picture has a face to fix. Set Face fix to Off to upscale faces like the rest of the picture.
Yes. An AI trained on photos redraws letters only a few pixels tall as strokes that were never there. So every picture is also checked for text (PP-OCRv4, a 5 MB model), and its small lines, such as the prices on a menu or the fine print on a label, are enlarged smoothly and sharpened instead, so each letter keeps its own shape. Large text is left to the AI, which draws it cleanly. Text the blur has already wiped out cannot be brought back.
The upscaler first checks how much detail the picture really has. A photo that is blurry or was already enlarged is reduced to its true sharpness before the AI redraws it, which gives cleaner edges than feeding it the blur.
No. The model downloads once into your browser and the picture is enlarged on your own device, on its GPU when it has one. Nothing is sent to or kept on our servers.
Up to 16 megapixels (for example 4096 x 4096). Larger asks are reduced to fit, and the size is shown before you download.
Yes. A PNG with see-through parts keeps them in the upscaled PNG. JPG downloads put them on white.
We keep adding free tools for shops and restaurants. Leave your email and we'll tell you when the next one is out.
We only use your email to write to you. See our privacy information.
