ccufcc

Has anyone experimented with secret annotations in prompts?

In older SD models, we could include things like \[some message::0\] in prompts — text that would be ignored by the model but still included in the raw input. This allowed us to embed metadata or trigger filters without affecting the image generation.

For example, we used \[ContentLabel::0\] to signal age restrictions. Our filter system treats this as a PG-13 banned prompt keyword, so users could self-label their content by adding it.

However, it seems that in newer models, this trick no longer works. Possibly because prompt weighting syntax like :: is no longer interpreted the same way — or ignored entirely.

Did anyone find a working alternative? It's important for us since we rely on this system to label and filter user submissions.

Here's my container or iframe for multiple plugin or link

...
  <div id="chatIframe4" style="width: 100%; height: 100%; overflow: auto; zoom: 0.55; border-radius: 0.25rem;" hidden>[image(settings.galleryOptions)]</div>
...

Only the [image(settings.galleryOptions)] would be stuck in initial load in color scheme; the [commentsPlugin(settings.defaultCommentOptions)] ; [settings.introMessage] ; perchance uploader or private gallery would sync with the gen theme toggle.

Before this I use long src url methods and it could have theme updated too!!

What's the trick here?

There have been many banned users—whether from chat or gallery—who claim they never committed the violations in question. Maybe they’re not telling the truth, but sometimes it really seems like they share the same IP ID or are using the same VPN. I don't know if that's actually possible, but I tend to believe them.

This makes the need for individually deleting or hiding gallery images even more urgent. We shouldn't ban everyone affected just because of someone else’s actions.

Heads up:

Files will soon no long be accessible via the old user-uploads.perchance.org domain unless they're programmatically requested from a generator. I.e. images used as backgrounds for generators, JS files imported, etc. should all continue to work without switching to the new domain (though I still recommend that you switch to the new domain). But if you want to view your file in the browser (i.e. by pasting the URL in your browser's address bar), then you'll need to use the new user.uploads.dev domain.

now should the old custom emojis we uploaded be safe to use in user-uploads.perchance.org domain, if only used in comment plugin?

or should we start transferring them to the new domain?

The dev change the file uploader url from

user-uploads.perchance.org/file

To

user.uploads.dev/file

But the pfp authentic validation didn't change it's rule

Now people are prohibited from uploading new pfp

I wonder if new emojis will suffer the same fate

I've noticed that certain terms or tags are causing rendering issues with the new model. The outputs are highly unstable and inconsistent—beyond what I would consider normal variation.

This doesn't appear to be due to new interpretation logic or prompt strategy shifts. Instead, many of these generations look glitched, underprocessed, washed out, or as if rendering was prematurely stopped. The saturation is often low, and overall image quality degraded.

I suspect that some of these tags may be acting like "stop codons", halting generation early—possibly similar in effect to using guidance_scale = 1.

From my testing, the problematic tags seem to fall into two groups:

Furry-related terms: furry, fursona, anthro, etc.

Illustration-related terms: drawing, line work, cel shading, etc.

It’s possible these tags are being masked or diluted when mixed with stronger or more stable tags, which may explain why some prompts still produce acceptable or mixed results. However, when multiple of these unstable tags are combined, the generation almost always fails—suggesting a kind of cumulative destabilization effect.

By contrast, photography and painting-style tags remain mostly unaffected and render normally.