AppStore Version Help Page
Thanks for downloading waifu2x-ios from App Store!
- Tap "From Photos" to select images from the album, which needs access to your photo library.
- Tap "From Files" to select images from the file browser.
- Use share menu in other applications.
- Use drag & drop on iPad.
All of the above methods support multiple selection.
Since v3.4.0 there is no more difference between single and batch mode. DO NOT pick too many images at a time since it will significantly heat up your device and drain the battery.
Image source: https://www.pixiv.net/member_illust.php?mode=medium&illust_id=60888049
Set parameters for model that you like to run. You can add more images to the queue by dropping images into the image view on the left on iPad.
Image source: https://www.pixiv.net/member_illust.php?mode=medium&illust_id=49017180
Wait for it to complete. Since v3.4.0, images WILL NOT be automatically saved.
It is NOT RECOMMENDED to continue if you receive warning about memory. All progress WILL BE LOST if the process is interrupted either manually or automatically.
After all processes are completed, you can manually save resulting images to your photo library (just by tapping 'Save to gallery'). You can also tap Action to invoke system share menu, from which you can export resulting images to files. From Preview you can compare results and view more details of the images.
By default output images are in PNG format. You can change it to JPEG or HEIF(from v3.4.1) in Preferences. Please note that JPEG format saves a lot of space (about 90%) at the cost of image quality. HEIF saves even more space. JPEG format is not capable of saving transparency information. Image quality setting for HEIF is only applicable if you export to files. If you save HEIF to the album, iOS will further compress it internally.
You can use waifu2x:// to launch this app in other applications (e.g. Shortcuts).
Alpha channel on transparent images is scaled using Bicubic interpolation. By default it is done on GPU. However, if the image is too large for your device to process (generally more than 8192px), it automatically falls back to Core Image. If it still fails, CPU will be used. In this way, please wait patiently since such algorithm is much slower on the CPU.
Metal is used to do some acceleration. Every image should be expanded by 7px before being processed. By default it is done on CPU. You can manually enable the option in Preferences to speed it up on GPU.
It uses Convolutional Neural Network to process blocks of image data. The models are running on Core ML. For more details, please refer to waifu2x.
This app is free to use without advertisements. The code of core function is open-source. If you feel this app useful, please consider giving me a cup of coffee from the drop-down menu on the right top corner of the app!
If you find a bug or have an idea of improvement, you can use one of the following ways to report it:
- If the app crashes, you can enable crash report in Preferences. It will automatically upload crash reports once you restart the app after crashing.
- Send feedback mail in About.
- Create an issue in this repository.
- Designed by HansWanglin and me
- Japanese translation by plasmas
- UICircularProgressRing
- Firebase
- BSImagePicker
- OpenCC
- Iconfont
- App icon background from Pixiv - Capura.L
- Pixiv
- Pixabay