Hey all, I just wanted to share something I’ve been finding useful. I know general sentiment towards AI (especially image generation) is negative, but I still wanted to share this in case other people find it useful.
There have been previous attemps to generate images for mnemonics, but they are all quite old and I don’t think they are that good. In recent months, ChatGPT’s image generation capabilities have vastly improved, especially with text within images. This allows us to generate images like the following.
It’s not perfect though with regards to the actual kanji itself within the image, which can sometimes be quite mangled, e.g.
However, I basically ignore the the kanji itself, and focus on the image, which seems to be doing quite a good job at helping with my progress so far. Since I’m currently around level 20, I’ve only been generating what I’ve been needing (so everything past level 19), found here:
Was wondering if anyone else would find these image mnemonics useful? If so, we could potentially make this a community repository, and that way we don’t have to re-generate a ton of images.
To get the images to follow a general pattern, I pre-generated the prompts, found here (sorry, this is my general WK repo, which was originally made for scraping the patterns of use / context sentences). I’m using ChatGPT Plus via the UI to generate images, but often run into rate limiting, so I generate around 4-5 a day, sometimes more like 20+, depends on how the rate limiting is treating me, because I don’t feel like paying for API usage.