If you go here: https://perchance.org/ai-text-to-image-generator, enter a prompt, click on the persona/chat button of an image, click generate persona, possibly edit the result, and then click chat with them, it creates a public link with the character’s image and the generated/edited text description. How long does that link persist?
You must log in or register to comment.
Should be indefinitely (no expiry). If you lost the link, and you don’t remember it or have it saved, then it is lost forever. If you have the character on the /ai-character-chat, you can regenerate a new link by clicking the link icon there.
I think this is inconsistent and confusing to the user, because in other situations the data is not saved:
https://rentry.org/perchance-ai-faq#does-it-save-the-data-that-i-put-into-it https://lemmy.world/comment/5709061
In the scenario I describe, the results of the prompt, additional input from the user (assuming they edit the character), and the image are all saved to the server and done so in a public and indefinite manner.
The two main issues I see with this are: One, the user’s privacy is not protected, if for example, their browser history is compromised–whoever sees their browser history also sees all their characters and the data therein (yes I know they shouldn’t have put personal information in them). Two, there isn’t any peer review process for the characters that are created. So an individual could create questionable characters/images and host a page linking to the perchance URLs of these characters. I know the team is fairly confident they can prevent illegal stuff from being created but it seems safer not to risk something like this happening. It seems like it would not help Perchance’s reputation.
This is just my opinion. Maybe my concerns aren’t a huge deal. But I just don’t see a major benefit of keeping them indefinitely that would outweigh even minor risks. Of course, maybe I am missing something big.
@[email protected] - I’ll ping the dev for other insights. But here’s what I think.
In the scenario I describe, the results of the prompt, additional input from the user (assuming they edit the character), and the image are all saved to the server and done so in a public and indefinite manner.
When creating a share link, the data of the character that you want to chat with is saved to the server, so that when going to another generator, that generator can retrieve the same data it needs to be. But only the required data (that is a JSON object compressed in gz) is the thing that is saved on the server. You can check the file that is generated through this URL:
https://user-uploads.perchance/file/<FileID>and on the character link likehttps://perchance/ai-character-chat?data=King_Spectrara~b28c24b93bf726eb1f850e12ee3f5d99.gzthe file id isb28c24b93bf726eb1f850e12ee3f5d99.gz.One, the user’s privacy is not protected, if for example, their browser history is compromised–whoever sees their browser history also sees all their characters and the data therein (yes I know they shouldn’t have put personal information in them).
Yes, if you navigate to the share link that is created, then that link is stored in the browser’s history and if anyone can get your browser history, can access the link. If you really want to be sure that your history isn’t accessed through the internet this option https://lemmy.world/comment/14698327 might be used. If you’ve added any sensitive/personal info on the character, then I think that would be on you. As long as no-one else knows the link or the fileID of the uploaded item, it wouldn’t be possible to retrieve the data saved on the server.
Two, there isn’t any peer review process for the characters that are created. So an individual could create questionable characters/images and host a page linking to the perchance URLs of these characters.
Yes, the uploading of the details to the server is automated and almost anyone can create share links to characters.
Why not just add the character data to local storage when you click “chat with them” on the ai-text-to-image-generator page and then just navigate directly to the ai-character-chat page?
Local storage is origin specific, meaning the local storage in
/ai-text-to-image-generatoris different from/ai-character-chatand there isn’t a way to pass data from one local storage to another. There are ways to pass data, but it involves using iframes and post messages.deleted by creator
These two pages share the same origin and can both access the origin specific local storage.
To see this, open https://perchance.org/ai-text-to-image-generator and https://perchance.org/ai-character-chat in two tabs of the same browser. In the first, click F12 to open dev console and paste localStorage.setItem(‘perchanceExample’, ‘iSeeYou!’) and press Enter. You will see it appear in the local storage of the second page.
Perhaps I’m not understanding what you’re saying. Or maybe am I, and it IS possible to do it the way I suggested but there is some other valid reason for not doing it like this that I haven’t thought of. The point is, it seems like you don’t really need these zip files and the fact that you have them anyway and are holding on to them indefinitely without what seems like a valid reason seems sus to me.
Each perchance page has its own local storage, which is by having a different domain. See this previous post. Then, each page is embedded on the main page, which is the top frame. By going to the Dev Console and setting the local storage there, it would be reflected on the top frame.
You can test by adding this to the HTML panel (bottom right panel after clicking ‘edit’) then clicking ‘reload’ on the bottom of the preview:
<script> localStorage.setItem("TEST", "a") </script>On one page and another, and it would be reflected to their respective domains on the local storage, instead of the top domain perchance.
I guess the purpose of saving the data is for ease of access and convenience for sharing instead of downloading the data - clicking import - selecting import file - confirm. Maybe the solution is to just add a ‘download’ button instead of creating a share link so the data is downloaded locally, but again, the steps of download - import - select file - confirm might be inconvenient for others.
As long as you don’t share the link or the file id. No one should be able to access the data from it.
@perchance - I’ll ping the dev again just in case.
