Releases: n00mkrad/cupscale
Releases · n00mkrad/cupscale
Build 20
Changes:
- Better CUDA fallback options; NCNN can now be used for preview mode and single-image upscaling
- Better seamless options - You can now choose between Tiled and Mirrored mode
- Interpolation value is now 50/50 by default
- Fixed bug for Joey's ESRGAN where the tile upscaling progress bar wouldn't be updated correctly
- Various UI/UX improvements (windows are no longer forced on top)
- Various logging & error handling improvements
Build 19 Hotfix 2
Changes:
- Fixed problem with "Advanced" model mode (outdated esrgan script)
- Added error handling for Nvidia API (won't throw an error on AMD hardware now)
Please go to the Settings and do a clean re-install of your resource files if you are coming from Build 19 or Build 19 Hotfix 1!
Build 19 Hotfix 1
Changes:
- Fixed Ctrl+V to paste an image not working. Bonus: You now no longer need to click on the preview image for Ctrl+V to work!
- Fixed error handling flaw where you can't do batch processing unless a preview image is loaded
Build 19
Changes:
- Added "Advanced" model mode that allows you to chain up to three interpolated (or single) models
- Added VRAM monitoring in the main view
- Added an option to automatically reload the image before upscaling (on by default)
- Fixed post-resizing not working correctly
- Fixed a bug where "Save Current Merged Preview" would post-resize the image once more
- Fixed alpha not working with NCNN
- Some more fixes regarding alpha, resizing, and color depth
- More error handling, as always
Build 18 [Hotfix 2]
Changes:
- Fixed "Full Image" preview refresh not working
Build 18 [Hotfix 1]
Changes:
- Fixed installer not including the updated ESRGAN files
- The new alpha filling method now also applies to dropped input images, not only for processing
Build 18
Changes:
- Added Joey's ESRGAN fork as an alternative to esrgan-launcher. Supports seamless mode, faster alpha upscaling. Experimental.
- Added "Result Only" mode for copying/saving comparison images
- Fixed bugged alpha filling for images that are not 24-bit
- Fixed edge cases where batch processing fails if you save images with source format and the source file has a capitalized extension
- Some UI improvements (Version number in main view, ...)
Build 17
Build 16
Changes:
- Added "Keep Folder Structure" option for batch processing
- Added button to save the current merged preview (no need to re-upscale if you previewed the full image anyway)
- The model selection window now accepts Enter/Return as confirm key, and Escape as cancel key
- Fixed "Same As Source" output mode not working correctly
- Fixed flickering progress bar