YomiNinja v0.7.2

Changelog

• MangaOCR integration (Currently available for Windows and Linux).
• Initial support for macOS.
• Integration of Apple's Vision Framework OCR engine (macOS only, generally superior to PaddleOCR or MangaOCR).
• Auto OCR (OCR Templates): Monitors your screen and automatically runs OCR whenever it detects significant changes.
• Text-to-Speech (OCR Templates).
• System tray icon.
• Overlay Adjustment Option: You can now manually move or resize the overlay either from the tray icon or by pressing Ctrl+Shift+M.
• Additional PaddleOCR languages: Chinese (traditional), Latin, and Cyrillic.
• Improved compatibility with the latest Yomitan versions.
• An OCR button was added to the tray icon menu.

Known Issues

• MangaOCR currently uses the Paddle Text Detector (for detecting text position). An update is coming to add the Comic-Text-Detector to achieve better results with manga (v0.8).
• MangaOCR takes up a lot of storage space. I'll try to make it optional.
• MangaOCR and PaddleOCR are not available on macOS. MangaOCR will be added soon (v0.8). I'm not sure if PaddleOCR should be added though; the macOS built-in OCR engine is superior.
• On macOS, before updating the app, you need to manually remove the app permissions (accessibility and screen recording) in the macOS privacy settings.
• Problems with multi-monitor systems on Linux and macOS.
• Potential issues with fullscreen applications on macOS.
• On macOS, the app might crash when exiting. You may have to force quit.
• The Apple Vision OCR engine doesn't work with vertical text. It seems like it was added recently in macOS 14 (I'm working to fix it).
• VOICEVOX TTS voices don't show up (fixable). You can use VoivoClip in the meantime.
• The app might fail to fully load on the first launch. However, it should start without issues on subsequent launches.
• Issues when signing in to jpdb.io using a Google account.
• Changes to PaddleOCR settings might not save.

Important Notes about Auto OCR

• The menu becomes available when selecting a region from the OCR Template screen.
• Make sure to manually select a capture source from the home menu when using Auto OCR.
• Currently only supports one region at a time.
• When the capture source window position or size changes, press the OCR hotkey to reposition the overlay.
• This feature is experimental, so avoid using it with Google Cloud Vision.
• It can be adjusted per region for more flexibility.
• Motion Sensitivity is the most important parameter. Try higher values for Visual Novels or RPGs and lower values for other types of games. Avoid changing other parameters unless necessary.
• I'm working on an option to let you track changes more precisely, especially in cutscenes.

Getting Auto OCR Working Step by Step

1. From the home screen, select a window as the Capture Source.
2. Choose an OCR Template.
3. Select the region containing the text you want to monitor.
4. Enable Auto OCR.
5. Adjust Motion Sensitivity.

Persona 3 Reload

Motion Sensitivity: 77
Number of Frames: 8
Capture Maximum Framerate: 3

Note: You can enhance your manga reading experience by using Google Lens and reducing the "size offset" of the bounding box. This option can be found under Settings > Overlay Appearance > Bounding Box. To log into jpdb.io, you can click on the "jpdb settings page" link from the JPDBReader extension settings screen.

Getting Started With YomiNinja

Support YomiNinja

Download

Windows:

• Windows x64

macOS:

• macOS x64

Linux x64:

• AppImage

• Deb

• RPM

• Pacman

YomiNinja v0.6.4

Changelog

• Google Lens integration.
• Google Cloud Vision integration.
• Individual character positioning for accurate text overlay (currently exclusive to Cloud Vision).
• Expanded PaddleOCR settings menu with additional parameters.
• Built-in Yomitan and JPDBReader extensions.
• Global hotkeys for toggling the overlay and copying extracted text.
• Dedicated global hotkeys for each OCR engine.
• Edit extracted text with Ctrl + Double Click, allowing for manual corrections.
• Enhanced overlay customization options.
• Option to automatically hide OCR results upon focus loss.
• Now you can change the installation directory.
• Added an AppImage option for Linux users.

Find more details in this post
Short demo video

Getting Started With YomiNinja

Note: You can enhance your manga reading experience by using Google Lens and reducing the "size offset" of the bounding box. This option can be found under Settings > Overlay Appearance > Bounding Box. To log into jpdb.io, you can click on the "jpdb settings page" link from the JPDBReader extension settings screen.

YomiNinja v0.5.1

Version 0.5 and 0.5.1 features:

• Chrome Extensions Manager.
• OCR Templates.
• Google Translate.
• Option to control which window is displayed when the text is copied (Overlay Behavior).
• Option to enable a custom mouse cursor for games that hide the cursor (Overlay Appearance).
• Options to control overlay font size and letter spacing (Overlay Appearance).
• Main window zoom controls.
• Improved overlay rendering.

Find more details in this post

YomiNinja v0.4.1

Version 0.4 and 0.4.1 changes:

• Pop-up dictionary usability improvements (automatic click-through and scroll).
• Option to show overlay without stealing focus.
• Fixed initial startup issue.
• Linux support.

Visit this post for more details

Yomi Ninja v0.3.1

Version 0.3 and 0.3.1 changes:

• Built-in popup dictionaries.
• Extensions toolbar buttons.
• Copy text on click. (only with Click-through disabled for now).
• Automatically handles CPU incompatibilities.

Visit this post for more details
v0.3.0 changes

If you're experiencing slow OCR processing after updating to this version, find the resolution in this issue.

Yomi Ninja v0.2.1

New features:

• Option to invert captured image colors.
• WebSocket support for texthookers. Port: 6677
• Option to alter the OCR inference runtime.

Visit this post for more details.

Note: If the OCR is not returning results, there's a new "Inference Runtime" option in the "OCR Engine" settings. Set it to "ONNX CPU" and restart the OCR Engine.

Yomi Ninja v0.1.0

First release features:

• Text extraction from the entire screen or specific window.
• Automatic text copying.

Recognition languages:

• English
• Japanese
• Chinese
• Korean