New Features

• The settings is re-designed, it is now more elegant
• The app would now reminder your last choice of parameters
• The image models are provided by default
• Now you can choose your video / image output format and codec
• Now in the memory only mode, you can choose to encode the intermediate frames in heic to reduce memory usage

UI Improvements

• Fixed the freaky issue where the alignment is wrong
• Adjustments to make the views more compact

Localizations

• Localized into Chinese (Simplified & Traditional)

Optimizations

• Some minor memory optimization
• Waifu2x performance improvements
• Redesigned the progress reporting workflow, making UI simpler, and introduced some bugs (maybe)

Model updates

• Updated the pre-installed models to the latest version

Bug fixed

• Fixed UI Issues
• Fixed the issue where the file name for images is not preserved
• Fixed Waifu2x issue where it may quit unexpected on intel-based Macs
• Fixed the issue where the app may crash unexpected if the output image is extremely large
• Fixed the error where the progress cannot be shown.
• Fixed the async issue mentioned in Release 4.1.3
• Fixed the error considering the grid view
• Fixed the extra memory allocated issue (The memory was not freed!)
• Fixed the concurrent perform memory issue (finally!)

Clean ups

• Removed the redundant Welcome View
• The structure of source code is reorganized, making the developer satisfying
• Removed the None destination option, it is confusing to use!

Language improvements

• Adopted the latest Swift Syntax (Swift 5.7).

• Added noneDestination

• Restructured Source code, making the developer happier

• Some minor memory optimization

• Fixed the issue where the app may crash unexpected if the output image is extremely large

• May have allowed installed models to run on multiple GPUs.

• The LazyVGrid is working again.

• Updated the pre-installed models to the latest version

• Fixed the model settings view

• Now you can choose your video / image output format

• Waifu2x performance improvements. Note: From now on, it would only check 10% of the pixels for checking whether an image contains alpha. Please inform me if you encounter any issue caused by this.

  • Now a standard 1080p image takes only 464ms with waifu2x anime scale level 2, on M1 Max.

• Finally fixed the freaky issue where the alignment is wrong
• A few adjustments to make the view more compact
• Fixed UI Issues
• Now TTA defaults to false. You can change this in preference.

Known Issue:

• LazyVGrid is not working properly.

5.0 Beta 7

• UI Improvements
• Removed the redundant Welcome View
• Removed some test UI which are only available in MacOS 13 introduced in Beta 6
• Fixed the issue where the file name for images is not preserved.

5.0 Beta 6

• Fixed localization issues.
• Fixed Waifu2x issue where it may quit unexpected on intel-based Macs.
• Introduced some new UI for Macs running MacOS 13, which is currently in Developer Beta.

5.0 Beta 5

• Localized into Chinese (Simplified & Traditional)
• Minor bug fixed
• This version is signed, so from now on your Mac should not inform you that this software is from unidentified developer.

Per @blacklein's request, from this release, the image models are provided by default. However, you would still need to install the frame models by yourself.

• Tiny UI improvements
• The structure of source code is reorganized, making the developer satisfying.

5.0 Beta 3 The app would now remember your last choice of model.

The second beta of 5.0 Fixed the error where the progress cannot be shown.

The Beta version of 5.0 This version uses the latest Swift Syntax (Swift 5.7).

You will need to re-link models as I was putting them in the wrong directory.

• Rewrote core to adapt the latest await / async technology

• Thanks to @sclsj, bug fixed and added the option to store intermediate files into memory.

• To make it easier for you to download / update this app or its models, the new versions no longer contain their packages into the app. Please download them from their authors and keep them somewhere where the app is allowed to access, for example, your downloads folder.

As the core was written, there may be potential bugs, please inform me if you find any bugs.

The new version is also observed to have some issues that may cause you unable to enlarge videos.

• When enlarging videos with Memory Only: completion handler was called inside Task { await }, hence, even after the app says it's done, you may need to wait until the audio and video were merged. When enlarging multiple videos, the next one may start before the previous one finishes.

• When enlarging videos without Memory Only: The success of enlarging depends on the speed of reading image from disk. If the reading is slow, it is more possible to succeed. I tested on a MacBook Pro with M1 Max chip and MacBook Pro with i5, all of which works fine.

This was caused by requestMediaDataWhenReady, which is required to merge images into a video. This particular method runs the callback multiple times on a mediaQueue iff the writer queue has finished writing a frame. The writer queue is hidden, hence it is not possible to observe whether the writer queue is occupied, and the enclosing function would return once the mediaQueue is free.

Please inform me If you find any workarounds.

Note: This issue is resolved in the next version by withCheckedContinuation.

• Improve speed of waifu2x (back to the way it was)

• Fixed the bug where enlarging videos causes it doubled the speed

• Various bugs fixed
• Fixed the critical bug that causes the failure of interpreting frames
• Updated rife and realcugan

Now you can compare among different models

Added the support for various new models Note: From now on, the app would be large. If you only prefer Waifu2x_Caffe and dain-ncnn-vulkan, you may consider using the previous releases.

Added support for realsr-ncnn-vulkan.

• UI improvements
• Bug fixed

Various bug fixed.

UI improvements and bug fixed.

• Now you can interpolation video frame

• Bug fixed
• New feature: now you can change output path

• Bug fixed

• Bug fixed

• bug fixed

• Adds the support of comparing images.
• bug fixed

The new version added the ability of using Metal to accelerate results.