AI image upscaler

Upload the image, choose 2x or 4x, and press Upscale the image. Make it with any DreamBench plan: the result appears beside the original on the page when the run finishes, usually well under a minute, ready to download as the full-size file.

No. It runs a super-resolution model that works from the pixels you upload: edges are sharpened and textures rebuilt from what is there, so a face stays the same face and text stays the same text. Face retouching is switched off on purpose; nothing is redrawn.

A slightly soft photo looks crisper, because edges and fine texture come back sharper at the larger size. If your image is still blurry after upscaling, the blur was in the original: a photo out of focus or shaken when it was taken has no sharp edge to rebuild, so it comes back larger with the same blur.

Twice or four times the width and height of the picture the run starts from, up to 4096 pixels on the long edge; a picture larger than 2048 pixels is scaled to 2048 first. At 4x, the model enlarges a picture of up to 1024 pixels 4x, and a larger one 2x, which is then resized to 4096. There is no 8K, which is 7680 pixels wide.

DPI, or PPI, is a note in the file that tells a printer how many pixels to put in an inch; it adds no detail. What upscaling changes is the number of pixels, which is what decides how large a picture prints sharply. Set the DPI your print service asks for when you order.

Photoshop has Generative Upscale, which enlarges 2x or 4x and can redraw detail, and Super Resolution in Camera Raw and Lightroom, which doubles a picture; Canva has an upscaler in its editor; Upscayl is an open-source desktop upscaler for Mac, Windows and Linux. This page works in the browser on any phone, tablet or computer, with nothing to install.

No. It upscales one still image at a time, and DreamBench has no video upscaler. A frame saved from a video can be upscaled like any picture, and DreamBench's video models make clips at their own resolutions in the composer.

Upscaling an image does not reduce its quality: no pixels are taken away. A plain enlargement spreads the same pixels further and looks soft; a super-resolution upscaler sharpens edges as it adds pixels. It cannot add detail the original never had, so blur and heavy JPEG blocks stay, only larger.

To upscale an image without losing quality, start from the original file and let a super-resolution model add the pixels: it sharpens edges and rebuilds texture from what the picture holds instead of stretching it. Upload graphics as PNG to keep them lossless, pick 2x if the picture is already large, and compare at full size.

Resizing stretches the pixels an image already has, so a bigger copy looks soft. Upscaling with a super-resolution model works out new pixels from the edges and textures in the picture, so the bigger copy stays sharp. Both keep the framing and aspect ratio; only upscaling keeps the sharpness.

Pick by whether the picture may change. A super-resolution upscaler, which this is, keeps the same picture and adds pixels: a face is still the face and text still reads. A generative upscaler redraws detail, which can look sharper but can change faces, lettering and texture. For a real person, a product or a document, keep the picture.

No. The width and the height grow by the same factor, so the framing and the aspect ratio stay exactly as they were: nothing is cropped, extended or moved. To change the shape of a picture, use the AI image expander, which adds new borders around it.

Yes, if it has scratches, tears or fading: the photo restoration tool repairs those, and the upscaler then enlarges the repaired picture. Upscaling first would only make the damage larger and sharper.

No. Every run needs a DreamBench plan; you create an account on the way, and the button shows the price in credits before you press it. A run that fails is refunded.

No. The upscaler takes one picture per run, and there is no batch upload. Run it again for the next one: each run is priced separately, and the button shows the price before you press it.

No. A DreamBench tool adds no watermark to an image it returns, on any plan, and the super-resolution model that runs this page embeds no mark of its own. The download is your picture with more pixels, nothing written or hidden in it.

Each run uses 7 credits from your DreamBench plan. Starter, $9.99 a month, includes 1,100 credits a month, about 157 runs of the AI image upscaler, plus AI video. A run that fails is refunded.

Yes: a free account and a DreamBench plan. The tools are part of every plan, from Starter at $9.99 a month.

A photo up to 40 megabytes: JPEG, PNG, WebP, HEIC or another format your browser opens. A photo straight from a phone is fine, an iPhone's HEIC included: the page scales a large picture down to 2048 pixels on its long edge before the run, so the result is never larger than the page says.

Yes. Under the DreamBench Terms you may use what a tool makes for personal and commercial purposes, including advertising, monetized content and work for clients, with no separate commercial-use upgrade. The Terms don't guarantee copyright in a result or clear another person's rights in your upload, so only use photos you have the rights to.

No. DreamBench doesn't use your prompts, uploads or videos to train or fine-tune AI models, and any future training program would be opt-in.

It stays in your browser until you make a result. Then it is used to make that result and nothing else, and removed from the tool's storage after seven days along with the result. A result you keep in your library stays as long as you keep it.