Enlarge photos and illustrations 2×–8× in your browser. Logos can stay on Lanczos. No upload, no account.
On GPU browsers, photos at 4× run Real-ESRGAN general v3 (~5 MB). Illustrations, badges, and line art use RealESR AnimeVideo v3 (~2.5 MB). Other browsers use Swin2SR for photos and Lanczos for illustrations. 2× photos always use Swin2SR. It does not denoise, recover faces, or match desktop tools such as Topaz Photo AI. Graphic / logo mode is Lanczos resize plus light sharpen — no neural net. A 4× result cannot exceed 8,192 px on a side. A 1,000×13,000 infographic is shrunk before upscaling — export those from the design file, or split them into shorter sections.
This tool is best for upscaling product photos or illustrations at 2x to 4x. It isn’t meant to replace desktop tools that offer more processing power. On mobile devices, the output is capped at 2,048 pixels, so Safari and Chrome don’t refresh during the upscaling cycle. Very tall and wide graphic files like infographics or full-page screenshots cannot be enlarged to a full 4x scale. Desktop upscaling is capped at 8,192 pixels. These files will be shrunk first, so the result will be blurry.
Drop files or a folder here
or click to browse · paste from clipboard
Accepts .JPG, .JPEG, .PNG, .WEBP, .BMP, .TIFF · Up to 1,000 files
Drop your files
Drag or click to open your files in the upscaler. If you’re upscaling a large file, I’d recommend doing one or two per batch so it finishes faster.
Choose settings
Adjust quality, format, and other options to match your needs.
Click Convert
Images upscale in the browser through WebAssembly. Nothing is uploaded.
Download
Download files individually or grab all at once as a ZIP.
No. Files are upscaled in your browser. Nothing is sent to a server. GPU browsers download Real-ESRGAN v3 (~5 MB) for photos and AnimeVideo v3 (~2.5 MB) for illustrations; other browsers use Swin2SR for photos and Lanczos for illustrations. Models are cached after the first load.
Photo: Runs Real-ESRGAN v3 at 4× resolution (Swin2SR 2× as a fallback). Illustration: Runs Real-ESR AnimeVideo v3 to better handle line art, badges, comics, etc. Graphic / logo: Lanczos resampling + a tiny amount of unsharp masking. Auto-detect: Sends few-color / flat-patch images to be handled by Illustration.
Short answer: yes. Standard upscaling will result in blurry images without the additional processing. An image upscaler reconstructs the edges. It won’t be as sharp as desktop tools that use larger models and don’t have face-recovery passes or noise-stripping.
Honestly, it won’t match Topaz or any desktop upscaler, but based on my testing, it does a decent job at upscaling images and graphics without stretching and blurring them. It does not match them. Topaz Photo AI and Gigapixel run several full-precision models on your GPU — denoise, sharpen, face recovery, then upscale. Run this tool on a desktop for the best results; the mobile version has a memory limit.
The best balance would be the 4× option. If you need faster results, use the 2× mode. The 3× is slightly faster than the 4× and yields a similar result, since it uses the 4× model with Lanczos downsampling. The photo upscaler runs on Real-ESRGAN 4×. The illustration upscaler uses AnimeVideo 4×, then Lanczos, which is slower, creates larger files, and is likely to hit browser memory limits (especially on mobile).
Typically portraits that have no recovery pass, noisy (or low-light) photographs, or old scanned pictures. Using the Illustration mode can result in a halo around logos. If that happens, switch to the graphic/logo option. Tall or wide photos or illustrations over 8,192 pixels are pre-shrunk to fit the browser canvas, resulting in a blurred output.
Depends on the infographic’s size. Anything larger than 8,192 pixels will be pre-shrunk. Use images under 2,048 pixels for the best results.
A 500 × 500-pixel image upscaled at 4× becomes 2,000 × 2,000 pixels, or around 6 × 6 inches at 300 DPI, which is good enough for a small print. This upscaler will not produce Topaz-level outputs, but I try to max out the output to get as close as possible. Let me know what your results are by emailing me at [email protected]; I’d love to hear from you.
Technically, you can do up to 1,000 files. But for the best results, and to reduce waiting time, do a batch of 10 images per to see how it comes out.