A 64-bit Windows 10, 8. 1, 8, or 7. At least 1GB of available RAM when you’re ready to process the image. For the fastest possible processing, you’ll want to have an NVIDIA GPU with Compute Capability 3. 0 or higher. Check your GPU’s computing capability at https://developer. nvidia. com/cuda-gpus. If you don’t have an NVIDIA GPU, you can process images with your CPU. If your NVIDIA GPU supports CUDA, install the CUDA Toolkit update from https://developer. nvidia. com/cuda-downloads, and then the cuDNN installation package from https://developer. nvidia. com/cudnn. This ensures that Waifu2x takes full advantage of your GPU’s processing power during the conversion, as processing with your CPU can be kind of slow. Unless you speak Japanese, you’ll want to use Google Chrome or any other web browser that has a built-in translation tool.

First, if you don’t speak Japanese, right-click the page and select Translate to English (or your language) so you can read the page. Then, click the latest version number under “Releases” in the right column. Scroll down to click waifu2x-caffe. zip. If the download doesn’t start immediately, click Download or Save to start it.

Open your downloads folder. Right-click waifu2x-caffe. zip and select Extract all. Select a location for Waifu2x. There’s no installer, so you can just choose its permanent folder location here. Click Extract to unpack the files.

If you do not have Windows set up to display file extensions, you may not see the “. exe” part at the end. Just double-click the file called waifu2x-caffe to launch the application.

If you do not have Windows set up to display file extensions, you may not see the “. exe” part at the end. Just double-click the file called waifu2x-caffe to launch the application.

This updates the default output path, which will be a new file in the same folder with a similar name.

This updates the default output path, which will be a new file in the same folder with a similar name.

This updates the default output path, which will be a new file in the same folder with a similar name.

Denoise and magnify is best used if you’re working with a small low-quality image (like an icon or avatar) and want to make it larger and clearer. A combination of upscaling and noise removal may still degrade the image quality, although the results still blow away the competition. Magnify only will work if the image’s quality is high enough that you don’t need to reduce noise. This simply upscales the image without removing artifacts. Denoise only is good for when you just want to clarify an image without adjusting its size. Magnify & Auto Denoise magnifies the image and runs an automatic denoiser based on the file type. For example, with this option, PNG files will not be de-noised and JPG files will. [2] X Research source

The higher the level, the more you risk blurriness or an oil painting effect. If you’re not denoising the image, you can leave Level 0 selected.

The higher the level, the more you risk blurriness or an oil painting effect. If you’re not denoising the image, you can leave Level 0 selected.

To get a more precise scaling value, you can use simple math. Let’s say you have a monitor with a 16:9 aspect ratio and 1920x1080 resolution: In most cases, the image you are working with will not be as wide as 16:9. Therefore, to calculate the scaling factor, you’d divide 1920 by the image’s width. For example, an image of 1500x1000 would have a scaling factor of 1920/1500 = 1. 28. You could enter 1. 28 as the magnification rate. In the case where the image is wider than 16:9, you’d divide 1080 by the image’s height. For example, an image of 1000x500 would have a scaling factor of 1080/500 = 2. 16. Waifu2x always rounds down when it calculates the final image resolution. If your numbers are not round like in the examples, always round up in the third or fourth decimal place.

To get a more precise scaling value, you can use simple math. Let’s say you have a monitor with a 16:9 aspect ratio and 1920x1080 resolution: In most cases, the image you are working with will not be as wide as 16:9. Therefore, to calculate the scaling factor, you’d divide 1920 by the image’s width. For example, an image of 1500x1000 would have a scaling factor of 1920/1500 = 1. 28. You could enter 1. 28 as the magnification rate. In the case where the image is wider than 16:9, you’d divide 1080 by the image’s height. For example, an image of 1000x500 would have a scaling factor of 1080/500 = 2. 16. Waifu2x always rounds down when it calculates the final image resolution. If your numbers are not round like in the examples, always round up in the third or fourth decimal place.

If you have a CUDA-supported GPU, you’ll want to use it, as the processing time will speed up tremendously.

If you have a CUDA-supported GPU, you’ll want to use it, as the processing time will speed up tremendously.

If you have a CUDA-supported GPU, you’ll want to use it, as the processing time will speed up tremendously.

If you have a CUDA-supported GPU, you’ll want to use it, as the processing time will speed up tremendously.

If you have a CUDA-supported GPU, you’ll want to use it, as the processing time will speed up tremendously.

It may take some trial an error before you get the image exactly as you’d like.

If you have Windows and want a full-featured version of Waifu2x, see the Using Waifu2x-Caffe for Windows method. If the page loads in another language you don’t understand, most browsers will prompt you to translate the page once it loads. If your browser doesn’t translate for you, check out Google Chrome.

Websites that host high-quality anime images are Pixiv, DeviantArt, and Konachan (wallpapers). If you find an image in Google Image Search, make sure to go to the page that hosts the image and copy the link (or download the image) from there. Otherwise, you might be starting with a smaller thumbnail version of the image with poorer quality.

Websites that host high-quality anime images are Pixiv, DeviantArt, and Konachan (wallpapers). If you find an image in Google Image Search, make sure to go to the page that hosts the image and copy the link (or download the image) from there. Otherwise, you might be starting with a smaller thumbnail version of the image with poorer quality.

Websites that host high-quality anime images are Pixiv, DeviantArt, and Konachan (wallpapers). If you find an image in Google Image Search, make sure to go to the page that hosts the image and copy the link (or download the image) from there. Otherwise, you might be starting with a smaller thumbnail version of the image with poorer quality.

A combination of upscaling and noise removal will still slightly degrade the image quality, but it’ll still look much better than it would if you simply resized it or upped the resolution in an app like Photoshop.

A combination of upscaling and noise removal will still slightly degrade the image quality, but it’ll still look much better than it would if you simply resized it or upped the resolution in an app like Photoshop.

A combination of upscaling and noise removal will still slightly degrade the image quality, but it’ll still look much better than it would if you simply resized it or upped the resolution in an app like Photoshop.

A combination of upscaling and noise removal will still slightly degrade the image quality, but it’ll still look much better than it would if you simply resized it or upped the resolution in an app like Photoshop.

If you aren’t satisfied, try again after playing around with the settings.