Upscale
Enlarge a photo by a percentage you choose, from 100% up to 300%, while an AI model reconstructs plausible detail instead of stretching pixels. 100% is a real setting: it denoises and sharpens at the original size without enlarging. It restores detail a lens actually captured, not detail that was never there. A very large original is enlarged only as far as the tool's output limit allows, so an oversized request comes back smaller than requested rather than failing.
Console
upscale-imageNothing is charged for uploading, and nothing runs until you press Run. Sign in to load a file and get an estimate first.
Upload your image and choose how much to enlarge, anywhere from 100% (no size change, just cleanup) to 300% of the original width and height. The estimate moves with the percentage you pick.
A larger percentage costs more, because it is both more reconstruction work for the model and a larger file to store and deliver afterward — you'll see the exact price before you commit.
Run and download. The file returns in its original format at its new resolution, and a run that fails costs you nothing.
Questions
How is this different from just resizing bigger?
Ordinary resizing has no new information to work with, so enlarging stretches existing pixels and the result looks soft. This tool uses a model trained on real photos to infer edges, texture and fine detail that are consistent with what a higher-resolution capture would likely show, which is why it holds up across the whole percentage band where a plain resize visibly blurs.
Can it fix a genuinely blurry or out-of-focus photo?
Not reliably — it enlarges and sharpens detail that is already present, faintly or at low resolution, rather than recovering information a lens never captured. A motion-blurred or badly out-of-focus original will come back bigger but still soft, because there was nothing sharp underneath for the model to reconstruct.
Why does a larger percentage cost more?
A larger percentage means both more pixels for the model to reconstruct and a bigger file for us to store and deliver back to you — 300% produces nine times the pixel count of 100%, not three. That is real, growing work and a real, growing file, not a pricing rule, so the price climbs with the percentage you pick right up to the 48-megapixel output cap described below — past that point the file stops growing, so the price does too, and either way the number you see before you run is exactly what you pay.
What does it cost?
Six credits as a base, plus six per megapixel of the original, plus a charge that scales with how large the enlarged file itself will be — so, unlike a flat per-photo price, the percentage you pick changes what you pay. A two-megapixel phone photo costs around 20 credits at 100%, about 25 at the 200% default, and about 34 at the full 300%. Whatever you pick, the number you see before you run is exactly what you pay.
What size limits apply?
Source files up to fifty megabytes, one photo per run. The output itself is capped at 48 megapixels regardless of the percentage requested, and for a large enough original that cap binds before your chosen percentage does: past roughly 12 megapixels the 200% default is already clamped, and past roughly 5.3 megapixels so is the full 300% — the file comes back enlarged as far as the cap allows rather than by the full percentage picked, and the estimate reflects the size the clamp will actually produce, not the percentage alone.