PhotosUpscalerUpscale a photoUpload

Anime and illustration upscaler

Crisp lines for anime and art.

Enlarge drawings and screenshots with clean edges and flat colours.

or drop it anywhere here · paste works too

FreeNo sign-upPhotos stay on your device

A kingfisher photo 170 pixels wide, enlarged as blocky pixelsBefore · 170 pxAfter · 4× sharper

Got a whole folder?

Queue up to 500 photos at once

Drop one illustration. Illustration mode cleans compression hardest and sharpens least, so flat fills stay flat and line art stays smooth, and the result is saved as PNG.

Step 1 of 3 · Files

Drop one illustration — or a folder to queue.

A directory or a selection becomes a queue with one rule for every picture in it. A single photograph — dropped, picked or pasted — opens the review studio instead, with a before and after at 100%.

Pixels are read and worked in this tab; model files travel to it, never the other way.

What you are shown before anything runs

— and — minutes

files · tiles · milliseconds a tile, measured over real tiles cut from your own photographs a moment ago, and the name of the graph those numbers describe.

The blanks are blank on purpose. Nothing on this page quotes a duration it has not timed on the machine it is running on, which is why the estimate arrives after the set and not before it.

Model weights come from our model file server only when you press the control that names their size. Everything else — the runtime, the decoding, the inference, the archive — is served from this domain and executed here.

What this page does differently

Illustration mode is tuned for flat colour: JPEG artefacts cleaned at 70 and noise at 45 before enlarging, detail recovery at 10 after. Flat fills have no texture to recover, and every pixel of “detail” added to one is noise.

The result is written as PNG by default, so the clean edges are not re-compressed into new blocks. Transparency in a PNG source is enlarged arithmetically and recombined, so a cut-out character gets no halo.

The risk map matters more here than on a photo: a reconstruction network trained on photographs can draw grain or texture onto a flat cel. Any 16-pixel cell where that happened is lit, and the tour visits the strongest.

What it will not do

  • Pixel art is not this mode's job. A sprite with a hard palette should be enlarged by an exact integer with nearest-neighbour, which keeps every pixel square; a model and clean-up soften it. The studio says so when the source looks like pixel art.
  • It is not waifu2x. The engine here is Real-ESRGAN general x4v3 (or ESPCN), run on this device; the result is close in spirit, not the same network.
  • It does not redraw missing line art or fix a drawing's mistakes.

Questions about this

Is my artwork uploaded?
No. It is decoded and enlarged in this tab. The only download is the model file, once, from our model file server, after a press that prints its size.
Is this waifu2x?
No. It runs Real-ESRGAN general x4v3, a compact reconstruction network (BSD-3-Clause), on your device, with clean-up and sharpening tuned for flat colour. The model ledger lists its exact length and hash.
Why PNG?
Because flat colour and hard edges are exactly what JPEG compresses badly. PNG keeps the cleaned result as it is, and keeps transparency.
Does it work on AI art and Midjourney images?
Yes — they are illustrations as far as the mode is concerned. If the image has photographic texture, Photo mode may suit it better; the mode cards preview each on your own image.
Why does it warn about pixel art?
Because a model and clean-up turn crisp square pixels into soft blobs. When a source has few colours and hard steps, the studio says that integer nearest-neighbour is the right enlargement for it.

Next on this site

Transparent artwork and logos have their own page; the risk-map page explains what a lit cell means on a flat fill.

Photographic visual for Anime Upscaler
Anime Upscaler