Screenshots

Description

This app can scale and/or denoise images using Convolutional Neural Network (CNN). It is the iOS Core ML/MPSCNN implementation of waifu2x. Note that this app is not affiliated with the original waifu2x project.

No ADs. No usage limits. No personal data collection.

Highlight features:
1. Unlimited number of imported images.
2. Supports both SRCNN and CUnet models from the original waifu2x project.
3. Exclusive PAN models for anime-style illustrations (requires purchasing).
4. Custom scale modes for PAN models. You have options to scale images up to 128x, 1.5x, fixed size and more.
5. Supports multiple ways of importing images: the album, files, drag and drop, other apps.
6. Supports images with alpha channel.
7. Multiple output formats: PNG, JPEG, HEIF, WebP.
8. Multiple processing backends. You can choose between highest quality or highest speed.
9. Write output images directly to an external path (including network drives like SMB).
10. End-to-end support for 16-bit color and color spaces. Color accuracy will not be compromised after processing.
11. Highly optimized RAM usage. You can easily produce images that don't even fit inside your device RAM.
12. Supports deleting original images if they come from the album.

SRCNN and CUnet models are converted from waifu2x-caffe.

All the models are running on device, so it doesn't need to upload image to any server.

Source code of core function is available on GitHub:
https://github.com/imxieyi/waifu2x-ios

Usage of images in screenshots is granted by the original author Kurumi:
https://twitter.com/0115_kurumi

The primary app icon is illustrated by Kapuru:
https://twitter.com/KaPuRu_0410

What’s New

Version 5.0.1

1. Added support for saving images directly to a destination path (requires advanced setup mode). You can even save to a remote path (for example SMB).
2. Image quality from SRCNN Core ML is very close to MPSCNN on iOS 15. It's no longer worth trading speed for quality. Core ML will be automatically selected if your device has Neural Engine and passes the sanity check. If you encounter any problem please switch back to MPSCNN manually and consider sending a feedback report.
3. Fixed libpng compression option not working for PAN models.
4. Fixed compatibility with M1 Macs.
5. Fixed several other issues.

Ratings and Reviews

5.0 out of 5
6 Ratings

6 Ratings

Gorki jatt ,

Best

Thanks for porting for mac and ios

App Privacy

The developer, 宜 谢, indicated that the app’s privacy practices may include handling of data as described below. For more information, see the developer’s privacy policy.

Data Not Linked to You

The following data may be collected but it is not linked to your identity:

  • Usage Data
  • Diagnostics

Privacy practices may vary based on, for example, the features you use or your age. Learn More

Supports

  • Family Sharing

    Up to six family members will be able to use this app with Family Sharing enabled.

You Might Also Like