Release 2.9.0

Changelog:

• Output layers can now be exported together as a Photoshop PSD file (they also work in other programs that support psd, like GIMP!). Thank you @al3ks1s!
• Italian language support! Thank you @Fujicrow!
• Allow overriding what language was detected, to force the use of Tesseract with any language you have installed.
• Masks may now be any solid color, not just grayscale (without needing to resort to inpainting).
• Image export parallelized in GUI mode, making it several times faster.
• You can now load previous .txt or .csv OCR outputs and re-edit them in the GUI.
• Window layout and sizes are now saved between sessions.
• The main window stays interactive while reviewing cleaning output.
• When system or graphics memory is low, a warning banner shows up. If left unchecked, the program will likely be terminated by the operating system.
• The image cache is now cleaned upon closing the program, not just opening.

Hotfix 2.9.1:

• Fixed CLI code failing with restricted output settings. (GUI unaffected)

Hotfix 2.9.2:

• Fixed how it's determined if settings are applied. (CLI unaffected)

Hotfix 2.9.3:

• Workaround for having corrupted performance counters on Windows

Release 2.8.0

Changelog:

• Check and edit OCR results for the entire batch
• Offer a preview of cleaning results for the entire batch
• Offer Spanish translations (thanks to Series all)
• Improve error recovery in the GUI
• Improve the way boxes are drawn in the details view
• Add an "Ignore All" button when loading duplicate images
• CSV is now the default OCR output format

Hotfix 2.8.1

• Make dbus-python an optional dependency

Hotfix 2.8.2

• Replaced the Qt compressed resource system with simple file bundling (Fixing "failed to load" resource errors)
• Sort images by file path in OCR review mode, rather than the order they were added in
• OCR review table now wraps lines and is more keyboard-interactive
• Fix: OCR review ignoring embedded image rotation tags

Release 2.7.0

Most of the work here is from the new contributor @civvic. Hold on tight, he's cooking up some even wilder features in the OCR space soon.
We now support Bulgarian in the GUI thanks to @librafrog.

Changelog:

• Support English text OCR with Tesseract
• Support Metal acceleration on Apple Silicon
• Offer Bulgarian translations
• Added a new output "Raw Boxes" that shows the other half of the raw data output by the comic text detector.
• [Fix] Links not clickable in pop-up dialogs
• A few other minor fixes

Note: Other languages that Tesseract could read aren't supported because the text detector doesn't support them, which is required for OCR to work.

Hotfix 2.7.2:

• Bring back support for Python 3.10. (Unnecessary to update binaries, as they bundle python)

Release 2.6.1

All right, now I can focus on DeepQt after these few bugs got me side-tracked.

Changelog:

• Add a light background to image previews to visualize actual boundaries of transparent images.
• Add a new option in the image previews to copy the image path to the clipboard, so that you can simply paste it into other applications without needing to export it first.
• Update image thumbnails in image details tabs after every step, rather than waiting until all steps are done.
• Only show number of errors in log viewer if there were errors.
• Fix: inpainting outputs not written to output.
• Fix: denoising and inpainting failing to scale masks correctly when using a sub 100% processing size.
• Fix: thumbnails not updating in details tabs when running processing.
• Fix: glitchy button text when regenerating an output in the details tab.
• Fix: bundled themes have poor contrast between enabled and disabled buttons

Hotfix 2.6.2:

• Fix: Windows cannot load resource files #88

Hotfix 2.6.3

• Force dependence on PySide6.7.0 or later since the resource bundles aren't backwards compatible.

This only affects the pip package, so no new executables.

Hotfix 2.5.3

Changelog:

• Fix: Regression in denoising failing when saving output in some cases caused by changes in 2.5.0

Windows exe coming later

Release 2.5.0

This is the final release for the foreseeable future as I shift my focus back to DeepQt.

Changelog:

• [Linux, Windows] CLI can now show desktop notifications about starting and finishing a batch when using --notify
• [Dolphin] You can now let Panel Cleaner (un)install a context menu to run the CLI when installed with pip, so that you can have images or directories cleaned without needing to open Panel Cleaner at all.
  (Support for Windows complete but not working in time. Pull Requests for other file managers welcome.)
• [Windows] Don't bundle everything into a single exe, resulting in massive time saved on startup.
• You can begin a GUI session with files already in the file table by simply listing them in the command line parameters.
• Fix: Denoising causing gray edges on bubbles in edge cases
• Fix: CLI not respecting the profile about inpainting being enabled

Note: only installing with pip will support CUDA.

2.5.2 Hotfix:

• [pip] pcleaner-gui didn't parse the command line arguments to display files on startup.

Release 2.4.0

Changelog:

• Experimental: support inpainting to clean boxes that couldn't be cleaned otherwise
  Note: inpainting is disabled by default, enable manually in the profile.
• Make the threshold for merging boxes configurable
• Changed masker analytics to show actual mask thicknesses, not the index
• Warn users that opening two sessions at once will corrupt the cache
• Make which output is selected in the details view clearer
• Fix: File table headers not spaced correctly
• Fix: Resetting a profile option shifts focus to the next option
• [Windows] Fix: magic not found upon pip installation

Note: only installing with pip will support CUDA.

Release 2.3.0

Changelog:

• Added a new details output: Fitment Quality. This will let you see what the standard deviation of the chosen mask is, if any.
• Analytics output won't be cleared automatically anymore, instead it can be done manually.
• Analytics output can now be edited by the user, so you can add your own notes while comparing effects of profile changes.
• Vastly improved error reporting, now enabling all users to easily see and include session logs (with username redacted for privacy).
• Added a new profile field to include notes about that profile.
• Clarified some profile settings.
• Support the XDG_CONFIG_HOME and XDG_CACHE_HOME environment variables on non-Linux platforms too. This will let you change where cache and config files are stored.
• Fix: Analytics mislabeling the largest mask as the box mask, if there was no box mask chosen.
• Fix: Flathub changelog not showing up

Note: only installing with pip will support CUDA.

Release 2.2.0

Changelog:

• Support translating the app
• Added language: German
• Fix: Cannot parse download progress from huggingface for OCR
• Fix: Image details view using wrong render mode on first load
• Fix: Overlapping boxes causing duplicate OCR text

Release 2.1.4

Minor visual UI enhancements.

Note: The exe and tar.gz bundles do not include CUDA libraries. See the readme on details for installing a CUDA-enabled version.

The .whl file is the same as downloaded by pip. You can install it manually too, if desired.