A mpv script to take screenshots and convert images with the waifu2x machinelearning upscale algorithm. This script is deprecated and might not work but I will leave it up here for reference and perhaps a future rework.
- Waifu2x and all it's dependencies
- The script uses the torch commands you can find in above link, test waifu2x out first with those
- This requires imagemagicks
convertcommand when using jpg output, test thatconvert image.png image.jpgworks - The script uses standard output of
pwdand%cd%(untested) for creating absolute paths if nececcary
- edit all the settings in conf variable in lua to match your system
- open menu with waifu2x keybind
- choose between screenshot(any kind of file that produces a screenshot) or image convert(safe for at least png/jpg for now)
- choose upscale amount
- choose noise reduction level
- the image will be created in 1-10 seconds
- I highly suggest using easy crop by aidanholm if you want to take cropped screenshots
- Screenshot will include subtitles if they are visible, hide subtitles(default v) to not capture them in the screenshot
- you can bind your own keys for waifu2x if you do not want to use the osd-gui, ex.
alt+x script-message waifu2x-send screenshot 2x 2in input.conf. Available options are image type, rezise, noise reduction and silent: [screenshot | image] [2x | 0x ] [no | 0 | 1 | 2 | 3] [true | false].
- "waifu2x" - default CTRL+S
- "waifu2x-up" - default dynamic key UP
- "waifu2x-down" - default dynamic key DOWN
- "waifu2x-enter" - default dynamic key ENTER
Dynamic means that the keybind will be active and override that key only when the menu is visible. You can disable dynamic keys and force static ones in the conf variable in the lua. Change keybinds in input conf with KEY script-binding waifu2x
- Add a non local version that uses some web api
- Port and test for windows
- image convert tests for different filetypes
- cudnn testing
- refactor the messy code