YomiNinja v0.6 (Patreon)
Content
Hello everyone! YomiNinja 0.6 is finally here!
Version 0.6 brings support for Yomitan, Yomichan, and JPDBReader extensions, along with Google Cloud Vision integration and various other enhancements.
Summary:
- OCR Improvements
- Enhanced Extensions Support
- Overlay Improvements
- Upcoming Features
OCR Improvements
Google Cloud Vision Integration
YomiNinja now integrates with Cloud Vision, a powerful image analysis service used by Google Lens. It significantly enhances OCR capabilities, improving accuracy and enabling the detection of the position for each character. Check out these examples of YomiNinja in action with Cloud Vision.
Rain Code - レインコード
Another Code Recollection - アナザーコード リコレクション
While not perfect, it significantly improves the overlay, providing a "cleaner" and more immersive experience. I'm impressed with these results and will be exploring further possibilities to bring this level of precision to offline engines as well.
To use Cloud Vision with YomiNinja, you'll need a Google Cloud account and provide a service account key. It offers 1000 free requests per month, and you can check the pricing details here. Setting up an account might seem a bit complicated, so I'm considering putting together a guide soon.
Note that when using Cloud Vision, your images are sent directly to the Cloud Vision API without intermediaries, alleviating privacy concerns. If you want to do a quick test, there's an option to use the demo API, allowing you to try it out without setting up an account.
More PaddleOCR Parameters
The PaddleOCR engine settings have been expanded. With the added parameters, you can make adjustments to improve accuracy. For example, sometimes "っ" is recognized as "つ" and by increasing the "Text area expansion factor" this can be less frequent.
Result with default parameters
Result with adjusted parameters
Expanded Extensions Support
Support for Yomitan and Yomichan is finally here! This is a game-changer, as these extensions offer a lot of features, especially custom dictionaries, Anki integration, and pronunciation audio.
JPDBReader is now supported, providing useful features like furigana, jpdb.io integration, and a popup dictionary. An interesting and very useful feature is its ability to help you visualize which words you already know.
Overlay Improvements
The rendering of the overlay was improved and can now more precisely position the extracted characters. Also, now it offers many more customization options, like font weight, letter spacing, font outline, color opacity, and more.
Sometimes the OCR results have simple errors that could be manually fixed, so I added an editing mode. It's a simple but effective feature that can help you in situations where basic OCR mistakes are affecting the dictionary look-up. To use it, just [Ctrl + Double-Click] on the text you wanna edit.
Additionally, while the Auto OCR feature is still in development, an option to automatically hide the OCR results when leaving the overlay was added.
Hotkey Changes:
- Added "Toggle overlay".
- Added dedicated hotkeys for each OCR engine.
- Improved and reintroduced the "Copy text" hotkey.
- Support for mouse buttons as hotkeys.
Upcoming Features
Following the tied poll results for the next OCR engine, I'll be integrating Manga OCR. This will be great for people who prefer offline OCR options. Also, I want to implement the Auto OCR feature and explore adding support for Mac OS. Other features could be included in the next release as well, so I will create a poll to better define the priority list.
A huge thank you for the amazing support and feedback! This wouldn't be possible without you!
Also, a special thanks to:
- Everyone involved in Yomitan, Yomichan, 10ten, and JPDBReader development for creating such helpful tools.
- GameGengo for featuring YomiNinja in his video.
- GameGrammar and Japanese Quest for welcoming YomiNinja into their communities.