Generate and display mnemonic images on WaniKani.
Install TamperMonkey:
Install the Userscript.
Any images previously generated by a user will be displayed on their respective subject pages and during reviews.
If an image has yet to be generated, a button will be shown allowing you to generate one.
The API server code and the UserScript are hosted on GitHub here.
Here are some good ones I generated.
This is a jukugo word that uses the on’yomi readings of the kanji. You already know the on’yomi for 温, but not for 厚, so here’s a mnemonic to help you remember it:
The most mild-mannered Tofugu staff has to be onion こういち. It’s an onion that jumps in for こういち whenever he’s not at the office. He’s the most gentle soul you could imagine, never yelling at anyone. The ideal boss, really.
Mnemonic
Your moon storehouse is another storehouse you own. Remember how your regular storehouse holds zombies (from the kanji 蔵?). Your moon storehouse is on the moon, and it is filled with internal organs.
Mnemonic
A horse gets trapped in a small room with a stool and an insect, and together they create a very boisterous situation. The horse and the insect are both scrambling frantically to get out of the small room, knocking the stool around all over the place.
Interesting. Reminds me of the similarly named project, though I don’t know if that has seen any development in a long while.
On a different topic.
I know this would probably be a lot harder, but if there was a way to hook into chatgpt.com rather than the API, that would allow it to work for people using Plus.
I learned that sharing OpenAI API keys is against the TOS, so I modified the backend to use my own API key. I set a limit of $50/month, which will allow for 1250 images/month. If this script blows up I will figure something else out, but for now it’s on me!
Also, I don’t think there is a way to call the image generation API with a Plus subscription. I think the best you could do is a userscript that runs on the ChatGPT page simultaneously and posts the images, but then my server would have to verify that the images are legit and not malicious.
I can’t see how the way it was implemented could be considered sharing of keys, unless you were storing a person’s key globally, which is unlikely based on the way it functioned.
Each person only used a single key to do a particular task. The end result of that gets collected into something for a shared project, but there’s no direct connection between the use of the API and the shared project.
But if you wanna feel safe that’s fine, and the route you’ve chosen is admirable.
