So it’s been a while since the last beta, and the reason for that was that we realized we had waited WAY too long to do anything about our own new stroke order diagrams and that we needed to drop everything we were doing and come up with a first release of those pronto, in order to have a chance to test them - with particular concern for the stroke order testing feature - before launch. So we had to quickly drop what we were doing and spin up a whole process for developing those.
Anyway, I’m pleased to say that we just released have our first take on official Pleco-developed stroke order diagrams - 27,474 characters based on FZNewKai, comprising all of Unicode CJK Unified + Extension A and missing only 1000 or so characters from any Pleco dictionary. (we plan to add those post-4.0 too, but we needed to hit this number to match the coverage from our old Song-style stroke order diagrams feature)
We basically started with the MakeMeAHanzi approach of applying some tuned Computer Graphics 101 math to Kai font outlines to turn them into separate strokes and then assigning an order to them based on the stroke orders of their components, but we made this process a lot more efficient by training a small local machine learning model on the first few few thousand outlines we cut up and retraining it as we went along, so that by the end, like 90% of characters were being processed entirely automatically and all we had to do was review its work. We then built a bunch of interesting batch processing tools to find mistakes / outliers / etc and fix those by hand. We’re sure there are still a fair number of bugs, but we don’t think it’ll be noticeably worse in that regard than either of our previous stroke order data sets, and unlike those, we developed this one ourselves and so can make further improvements to it very efficiently.
We’re currently making these new stroke order diagrams free for 4.0 beta testers, but we do plan to charge an upgrade fee for them in the final release of 4.0 (though as with other updates beta testers who gave us useful feedback will likely be comped), in large part because we had already been planning to charge for access to the stroke order testing feature and this seems like a convenient place to do that rather than have two separate stroke-order-related upgrade fees.
As FZNewKai is a simplified-style font, we also wanted to offer a better option for people who prefer traditional-style characters (flipped 骨 box e.g.), and it turns out that the Taiwan Ministry of Education’s stroke order diagrams website offers its data under a Creative Commons BY-NC-ND license. We’re following the “BY” part of that license by acknowledging the source, the “NC” part of the license by posting the extracted original data here as a free download, and the “ND” part by posting only the original data, which we have wired up Pleco to work with directly rather than having to convert it to our own database format. Here’s the file:
moe-stroke-order-data-260415.zip
To use this in Pleco 4.0 Beta, simply copy this .zip file into Pleco’s storage area, go into the stroke order settings screen (gear icon in the stroke order popup), click on the option at the bottom to load alternate stroke data, then select the file. This will only work if you have our new Kai stroke order diagrams installed and activated, but you can configure the app to use these in preference to those or even to use them exclusively.
Along with this, we also added a flashcard list for the 2026 revision of HSK 3.0, and revamped our website and online store (which had been overdue for one ever since we started taking in-app orders on iOS and our web sales exploded). New HSK list will be in our legacy apps in a few days (and is up now in Flashcard Exchange), new stroke order will probably come to them soon too as the data format is the same and there’s no particular reason to wait.
Anyway, I’m pleased to say that we just released have our first take on official Pleco-developed stroke order diagrams - 27,474 characters based on FZNewKai, comprising all of Unicode CJK Unified + Extension A and missing only 1000 or so characters from any Pleco dictionary. (we plan to add those post-4.0 too, but we needed to hit this number to match the coverage from our old Song-style stroke order diagrams feature)
We basically started with the MakeMeAHanzi approach of applying some tuned Computer Graphics 101 math to Kai font outlines to turn them into separate strokes and then assigning an order to them based on the stroke orders of their components, but we made this process a lot more efficient by training a small local machine learning model on the first few few thousand outlines we cut up and retraining it as we went along, so that by the end, like 90% of characters were being processed entirely automatically and all we had to do was review its work. We then built a bunch of interesting batch processing tools to find mistakes / outliers / etc and fix those by hand. We’re sure there are still a fair number of bugs, but we don’t think it’ll be noticeably worse in that regard than either of our previous stroke order data sets, and unlike those, we developed this one ourselves and so can make further improvements to it very efficiently.
We’re currently making these new stroke order diagrams free for 4.0 beta testers, but we do plan to charge an upgrade fee for them in the final release of 4.0 (though as with other updates beta testers who gave us useful feedback will likely be comped), in large part because we had already been planning to charge for access to the stroke order testing feature and this seems like a convenient place to do that rather than have two separate stroke-order-related upgrade fees.
As FZNewKai is a simplified-style font, we also wanted to offer a better option for people who prefer traditional-style characters (flipped 骨 box e.g.), and it turns out that the Taiwan Ministry of Education’s stroke order diagrams website offers its data under a Creative Commons BY-NC-ND license. We’re following the “BY” part of that license by acknowledging the source, the “NC” part of the license by posting the extracted original data here as a free download, and the “ND” part by posting only the original data, which we have wired up Pleco to work with directly rather than having to convert it to our own database format. Here’s the file:
moe-stroke-order-data-260415.zip
To use this in Pleco 4.0 Beta, simply copy this .zip file into Pleco’s storage area, go into the stroke order settings screen (gear icon in the stroke order popup), click on the option at the bottom to load alternate stroke data, then select the file. This will only work if you have our new Kai stroke order diagrams installed and activated, but you can configure the app to use these in preference to those or even to use them exclusively.
Along with this, we also added a flashcard list for the 2026 revision of HSK 3.0, and revamped our website and online store (which had been overdue for one ever since we started taking in-app orders on iOS and our web sales exploded). New HSK list will be in our legacy apps in a few days (and is up now in Flashcard Exchange), new stroke order will probably come to them soon too as the data format is the same and there’s no particular reason to wait.
Last edited:
