AI Image Upscaler

How to use

1. Choose a picture. It lands in the well, and the plan beside it lists the size you'll get back, how many 256-pixel tiles the run takes, and whether they go to the graphics card or the processor, with the seconds that costs.
2. Press 2× or 4×, or press Exact size and fill in what you actually need. If the shape you asked for differs from the source, a row offers Auto, Crop to fill, Fit inside or Stretch, the well draws the frame with the picture where it will land, and a Border colour field appears when padding is involved.
3. Set the download format and quality and press Upscale. The working button counts the tiles with the seconds left. The result opens at 1:1 with a divider between the browser's own enlargement and the model's; Fit shows the whole frame, but drag the divider at 1:1 before you download, because an upscale judged at a third of its size tells you nothing.

About this tool

Enlarging a photo the ordinary way spreads the pixels you already have across more area, which is why a small picture blown up looks soft. The upscaler does something else. A super-resolution network reads the image in 256-pixel tiles and paints in detail consistent with what it can see, so edges hold together at 2× and at 4×. The two models were trained on pictures that had been JPEG-compressed and resized half to death, which is what happens to anything forwarded through a chat app, so blocking and ringing tend to clear up on the way rather than get magnified.

Next to 2× and 4× there's a third button that turns out to be the one most people want. Exact size takes a paper or photo size from the list, or a width and height you type in pixels, millimetres, centimetres or inches, converted through the DPI box that starts at 300. The fit decides how far the picture has to grow. The network only knows two factors, so it runs the smaller one that clears that and resamples onto the number, which is sharper than stretching straight there; a picture already large enough for the size runs no model at all and is simply resampled. Below that you choose PNG, JPEG or WebP with a quality from 1 to 100, and the DPI you set gets written into the PNG and JPEG file so a print shop reads the size you meant.

Good to know

• It rebuilds plausible detail. It cannot recover what the camera never recorded, so text too small to read stays unreadable and a face a dozen pixels across won't turn into someone you recognise.
• Speed depends entirely on WebGPU. A tile takes about 0.075 seconds on the graphics card and about 1.15 seconds on the processor, which is the difference between four seconds and a minute for the same photo. Once a run is under way, the seconds left come from the tiles it has already timed.
• Input is capped at 12 megapixels with WebGPU and 2 megapixels without. Past the cap the refusal sits where the plan's numbers are, Upscale is held, and the page names the size that would fit, with Shrink it in the Resizer beside the button.
• Asking Exact size for more than 4× works, but everything beyond the model's factor is plain enlargement and the plan says so. Detail stops improving at that point. Asking for less than the picture already has runs no model: the plan says it is already large enough, and the press only resamples.
• Transparency survives as PNG or WebP only. JPEG has no alpha channel, so a cut-out saved that way comes back sitting on a solid colour.
• It's a general photo model. Faces get no special treatment, small type gets no special treatment, and motion blur or a missed focus stays exactly as blurred as it arrived.