On my generators I have a feedback button in the bottom-right of the window. Now they are covered up by a "Privacy" button that never goes away. 🙄
A heads up: sign up, and while logged in you won't get ads anyway.
Like the ai character chat?
You know anyone can make their own AI text generator, right?
Could you remove that giant thumbnail perhaps? Might have to remove the link of the post, and just put it in the body to do that. It's just taking over the main feed pretty heavily unfortunately. I don't know why it works like that on here.
You need to @perchance@lemmy.world in a comment for them to get pinged, apparently.
My 2 cents... The way I'd like to see that implemented is, if the target gallery is changed you're allowed to click the "save to gallery" button again.
There is no such set of terms it understands. There's just the process of trying different terms, and seeing what happens. It's not a rigorous structured database. It's an amorphous blob of abstract meaning linking even more amorphous probability spaces of different kinds of images linked to words images were tagged with before it was trained on them.
I'd just google it. 🤷
Something to bear in mind though is that in general, the fewer prompt terms, the better. You can have gargantuan, monolithic walls of text as prompts and negative prompts... which will take more time for the engine to process and generate from, and is more likely to produce the weird artefacts you're trying to avoid.
Just something to think about.
Copy the address of the generator. Paste it here.
Save the evaluated prompt text first. Then use that for the image options object, and the display of the text in the page.
To evaluate a randomising piece of text like that, reference it like this: [output.evaluateItem]
To save that into a variable, do something like this: [final_output = output.evaluateItem] Have that run before referencing it using [final_output] in the image options, and in the page.
Tricky thing is, the AI is fed what it has said before so it can continue the thread. I'm guessing that's the issue--it's reinforcing itself, making that aspect stronger and stronger in its own prompting.
No idea how to fix it though 😅
So to avoid the potential of someone going to the generators page in the 5 second window in which it's public--the first and only time it ever will be--you'd make an entire new email address, and perchance account, and set that up in your browser so you can have both accounts open at once? You do you, but man that seems like a lot of work to avoid that very slim chance 😅
Are you asking about the Lemmy forum specifically? It doesn't just delete things at random or something like that.
You can find all your own comments and posts by clicking on your name (top-right) and clicking on "profile." Then you can filter by comments/posts etc.
They are generated. The whole image is generated. Hence the term "AI generated image." It's never just copy-pasted from a real photo.
Even when you use the name of a celebrity, that face is generated... to look similar to that celebrity. But it's still generated.
Are you planning on signing out every time you want to work on a version? And making a new version?
What I do while working on a new version is, I work on a private/unlisted copy of the generator. I find that a lot easier. And I can reuse it too. That other copy becomes the backup of the code. And I copy the code across to the public "real" version when I'm ready for the world to see. Just a thought.
I'm not an AI-chatter myself, but I am still seeing many reports of weirdness.
Helper functions to add and remove children from a Perchance list.
My newer generators are not having their screenshots generated. https://perchance.org/perchance-object-reference
Though this one works somehow: https://perchance.org/list-management-plugin
They're using the same formatting and scripts for the page as all my others.
Currently it's quite fiddly, making sure all the keys/names arrays are up to date so it still functions as expected.
If it used a set and delete handler it could handle all that internally. Or if there was some special methods we could use like .addValue() and it would figure out what kind of value it is--property, node, function--and handle it correctly.
Of course, I'm working on a plugin now to handle this for people. 😜
There's a lot of stuff that becomes real fiddly to do because they are not iterable... that would be real easy to do if they were iterable. I believe there's a proxy "trap" for iterating over properties somewhere in there.
Override the device's colour scheme using existing styles.
This plugin overrides the device's color scheme, but uses the existing styles. No changing of styles is needed whatsoever; it just works.
Most likely SUPER not a priority. But you know me 😅
https://perchance.org/prompt-style-tester#edit
The "go" button updates the entire page. As the options are generated this means they are re-generated, which means the chosen option resets, and you can only ever test the top option: Painted Anime.
Simple fix though: put a div with an id around the generations, then update that div only. That's all that needs updating anyway.
As we have ignorePerchanceErrors() and clearPerchanceErrors(), seems fair to have this function too, for public use.
Would be useful to be able to see the images just in the page so we don't accidentally delete the wrong one, stuff like that. I know we can click on the link, but this would just be a nice feature to make it easier.
A primer on how to code using the Perchance language. Learn as much as possible as quickly as possible.
Another day, another learn...ing...s.
This covers everything from how to start editing/making a generator, to writing perchance code. (The more web-y stuff are covered in my other tutorial.)
Seems to be related to this: ___htmlToElements(). It takes a string which is the text, ignoring the fact that HTML entities are not normal characters and should not be interpreted as normal HTML. For example, < will be interpreted as < which can open a tag.
Example: https://perchance.org/qzi72h52te#edit
So if an element contains any perchance code, and HTML entities, the HTML entities are essentially lost and become real characters. Which isn't what you want if you're trying to present plain-text code to the user.
Definitely does not work as expected at least.
For example, I'm linking to the Learn tab but that doesn't actually open the Learn tab.
No error message makes it to the display.
In the console there is:
Critical error in __createPerchanceTree: TypeError: Cannot read properties of undefined (reading 'startsWith')
And in the error messages there's just this:
Your generator's script seems to have errors in it. If you haven't recieved any other errors above this one which could indicate what went wrong, then this could be a bug in the Perchance engine. *etc...*
("recieved" should be "received" by the way 😜)
As seen here: https://perchance.org/p5bdhjbfv6#edit
So it still runs somehow? And accesses the grandparent somehow? Likely a bug with how escapes are stripped out--and perhaps recursively processed or something...
https://perchance.org/2p16eseroi#edit
Got a tricky situation here. If I do a .replace() on the html of an element (just some formatting stuff for code blocks), the update() is still referencing old text nodes. So it just doesn't care about the new ones that are there from the replace and they stay there. And it adds new ones that are actually updated.
Not sure what to do about that, honestly. Ideally I'd be able to do this replace before the initial update of code blocks is run. Is there some way of doing that?
The issue seems to only happen for code blocks? Maybe?
Based on the newly added explanation of the public API calls themselves, I've made a little plugin for using those APIs a little easier, along with the usual demos and explanations.
If I have padding on the body, it doesn't do much. Because there's height:100% on the body by default, which is actually incorrect and causes this issue for reasons. As seen here: https://perchance.org/o0demyuvpj#edit Whereas without any default styles the padding would work as expected.
It should be max-height:100% instead, which would not have the problem. Could I request that be changed?