Tattoo artist all images refused

I can’t get it to create “beautiful” characters or basically any other descriptions of people or scenes… even blood as a solo prompt?.. won’t even give me a donor bag or blood vessels cells or anything? Do people not have body parts and nothing is beautiful or ugly now? What in the actual hell is going on

Welcome to the community @takksplace

Can you share the prompt you are using, so we can try to point you in the right direction? If we have a better idea of where you are trying to go, we may be able to help get you there.

1 Like

It just refused all “negative” imagery of any kind, and almost all positive in relation to people, and then I tried public figures and it makes like wooden dolls that look nothing like them…? it will do general things like “cosmic horror” but nothing specific to help speed up customer designs

We need to see your actual prompt and negative prompt in order to assist. Also resolution? Sampling method? CFG and iteration steps settings?

so I canceled my subscription because that’s the only reason I’ve had it for this time and it’s bricked

Sorry I am getting this confused with the Stability forums.

its saying i can only post one image at a time so hopefully this works… jeebs

I don’t understand. Can you see the screenshots of my prompt and it’s refusals?

Lol I have no idea what’s happening anymore

My apologies I meant no disrespect, I thought this was the Stable Diffusion forum and was taken aback by your images.

So your not really giving enough information to work with, paint it a visual picture. Also it helps to ask “Generate an image of…” so try “Generate an image of an crying infant child, wrapped in a soft blanket, being soothed by his mother.”

Oh you’re fine

And about the images…that’s the thing… it will probably do a crying infant getting comforted because it’s “nice” but a kid crying without context which is what I need to brainstorm… apparently not. And I do start with generate an image of… but I found that within the same thread it understands a simple “one more please” or “now this” etc

As an example. It understands what I’m asking for and even when I just throw out another name… its just so bricked that nothing’s possible

Ahh okay, yeah I’ve mostly been using Google Gemini, which unfortunately isn’t producing images of people right now period. I’m going to set aside time to mess around with Anthropic’s new Claude model to see what the fuss is all about. Your might have better luck there.

So if you’re using image generation provided with a subscription based model, you kinda have to go with what you’re given. If you have the hardware in your PC, I’m running a 10GB Nvidia RTX 3080 GPU but you could probably use most 3000 series cards just fine, you can download and run stable diffusion locally. This gives access to a versatile generative model with many options to fine tune your image to match your initial vision. Stable Diffusion 3 comes out soon (hopefully by end of next month) and should perform better/faster on less powerful hardware. But most of these online image generators are guard railed into oblivion.

1 Like

hmmm. yeah I didn’t realize I could run one locally or more importantly that a local version would have different guard rails. I’m absolutely switching to that route especially if 3000 series will get me there.

cheers for that aaron

Not problem I recommend the “forge” setup package I guess from hugging face? I’m sure if you search “forge stable diffusion” you find it. There’s several installers for it, and I spent an entire day pissing around in cmd.exe trying to download libraries and whatnot. The forge one just runs a batch file and automatically installs it.

Also don’t forget go to StabilityAI’s website and sign up for the wait list for the Stable Diffusion 3 early preview. Like I said it’s supposed to perform several magnitudes better on lesser hardware, and have next-level prompt adherence. Personally I can’t wait.