1. In what way would React improve performance? It's still JavaScript. I never used React, but isn't it possible to just add the script from a cdn? https://dev.to/kokaneka/react-development-using-just-a-single-file-5e65
2. Yeah, because afaik it is actually testing what it does while the website is blocked
3. My web dev knowledge is not elevated enough to understand what that does and why it needs to be sopported

Ok sorry, I'm not much into training models myself, so you probably know more than me.

at some point it was trained in some dataset that says it was trained by google

What if the model was trained on the same dataset that said it's trained by google, but it isn't actually a google model?

And what is a Cloudflare Worker and how do you know perchance is using it?

This is the title of a Lemmy post from two days ago. It's from the same guy here in the comments who said it's gemma...

Tell me you know nothing about LLMs without telling me... You can't ask it about stuff like that. It'll hallucinate something (especially since you mentioned google in the prompt).

768x768 is pretty new; the documentation is probably outdated. Forcing the resolution with (resolution:::768x512) is working for me. Where do you put it? Try putting it at the end of the Image generation prompt, and make sure it's not mentioned twice anywhere in the prompt.

That being said, the default is 768x768 unless "(resolution:::" is present or any of the following keywords are included: "portrait" or "selfie" for 512x768 or "landscape" or "wide" for 768x512 It will always prioritize the value of "(resolution:::" above any keywords

I think a link to the generator would be helpful

This is the only thing I found https://perchance.org/ai-character-chat?data=Fire_Alarm_Bot%7Ed98b51aee62731fb6a421b46aae1485c.gz

It actually got updated recently. It's now using Fast bigram-based approx token counter. Not an AI guy and no clue what that means, but it's not the old "count/3.6" method anymore :D

• When you are figuring out how to pull some piece of data out of an object created by a Perchance plugin, how do you figure out what the Javascript name of that object is?
  • Not sure I understand correctly. Do you want to extract a list out of a plugin? Then something like this should do the trick: "root.$moduleSpace["nested-plugin"].world". Or do you want to understand how to find it? I do that through inspecting the root object in the browser console
• Is there some standard way of figuring out which account maintains a Perchance plugin is if their name is not displayed on the plugin’s page?
  • Not that I know of

Now to your main problem: Afaik, perchance objects are static and don't save their state when they're evaluated. You have to inspect the DOM with JavaScript and traverse the parent elements to gather all descriptions. At least that's how I got it https://perchance.org/ah70cbqlqk#edit

I'm not really good at explaining, please ask if something is unclear. What I've done in the example is, I've overwritten the javascript onclick handler defined in the nested-plugin. This onclick handler is executed when expanding a header element. This allowed me to inspect the DOM that was generated through the nested-plugin. If the currently expanding element is a leaf, it traverses each parent header element and extracts the description of it and overwrites the content with all gathered descriptions.