So, I posted about this a bit back, but there have been some odd developments that may provide context and clues for anyone to assist in finding a solution.
So, for some reason, I can no longer save images to galleries (doesn't matter what generator I use either). While sure, I could always just download them, Perchance seems to learn what you want better based on your saved and thumbed up images. So, by not being able to do that, every time I use it-- and even every 30 min - 1 hr when it seems to refresh or something-- it starts back at square one, having forgotten all of the signals towards what I'm looking for from prior images and making the same obvious mistakes it seemed to have learned to avoid just moments prior.
However, every now and then, I can create a new gallery and images will save to the gallery for a few hours. But, inevitably, it will stop saving new images.
Keep in mind, I'm making sure to include the adult age of the models in the images and other elements to prevent triggering certain filters. Now, if I was simply no longer able to ever save images to galleries, that would be one thing. But the fact that I can occasionally start new galleries only for it to eventually lose the ability to save more images makes me wonder what else might be going on. A periodic loss of the ability to save images seems odd.
Not sure if there is anyone who knows what's going on or how to assist. And, I acknowledge I may just be speaking gibberish as I'm more trying to sus it out and intuit based on observation as opposed to actually understand how it works.
While the underlying image models do learn. The image generation that Perchance uses are continuously updated and trained on vast datasets by it's creator. This is why users may notice a difference in the quality of the generated images over time. For example, a model may be updated to better interpret a user's intent or to generate more diverse imagery. User interaction doesn't train the model. The tool does not learn from the prompts nor from the upvotes you personally enter. Providing feedback or generating many similar images does not cause the model to adapt specifically for you. The improvements you see are from broader model updates rolled out to all users. The user learns, which improves results. For a user, the "learning" process is about refining their prompting techniques to achieve the desired result.
with what you have said in response to point 2. just test it, open edit on the image gen find the line that says hideIfScoreIsBelow and make the number witch will likely be -1 change this to something like -15 and then hit reload (not refresh) check the gallery you post stuff to, swap it to recent / none /also want to point out cause i just thought about it. If you only do perchance on private browsing there is a feature which hides nsfw images until the generate button is pressed X amount of times, i cant recall the amount. do all of that see if they show, if they do not enter into admin mode. click the cog for the gallery and type the word "admin" press enter. if the images appear and they have a red box around them this means they have been banned or hidden by the admin of that page.
edit some words
Okay, so changing the score doesn't impact the retention of the images. And, the next day, the entire gallery disappeared. Also, if I hit refresh too many times, the gallery simply fades into the aether. I have also tried this with every browser configuration I can think of, and nothing has altered it. Likewise, admin mode showed no red boxes.