@Airathias None of your examples are using URPM, a few of them are using a merge of that model with a few others. It looks like you used UberRealistic Dreamy PFG Porn, which is a merge of URPM, OpenJourney and PFG.

I will try again and see if increasing the weight helps. I did put the triggerword at the beginning, and tried messing around with different samplers, steps, scale, none of it worked well, so I gave up as the results I was seeing were little different than the usual. I assumed the model was the problem.

@Airathias I'm confused by this, because when I copy everything exactly from what you are doing I get different looking images and the tongues are messed up. Now, I'm using a slightly older version of URPM, so that could account for it. But also, your images don't look like anything coming out of URPM, yours look like a cross between a photograph and a drawing, whereas mine are photorealistic.

So, either the newest version of Uberrealistic Porn Merge is much less uber realistic, or something else is different that I'm not aware of. I'm using automatic1111, on a colab.

Why are the pictures you posted not photorealistic looking? That could account for it, as I know that the novelAI model, which is trained on anime, can do tongues vastly better than other models.

@Airathias Ok, I figured out the problem.

I was using the Restore Faces function in Automatic1111. It turns out that Restore Faces does not play well with your textual inversion, and in fact completely messes it up. I don't know how common it is for people to use this, but anyone who does so will believe awaitingtongue works very poorly. Might be worth warning people not to use this.

If you are using Stable Diffusion WebUI like I am, it should go into this directory: '/stable-diffusion-webui/embeddings'

I do recommend to go to extensions in the webui and install the plugin for textual inversion first, otherwise it probably won't work.

@HaokiChan thanks bud!

@HaokiChan where can i find the plug in?

@HaokiChan you are a hero! I am I didnt't really even attempt to mess with pt files yet, had just been stashing them away and favoriting/thumbs-ping/bookmarking as I found some ones I liked. But I have to assume the key info you just gave likely would have been presented in a "not so clear" way, based on my experience.

Tending to take too much before I understand the last thing, is my issue, but now that is one less thing to go into blindly.. This whole LLM, AI ML, whatever, world has an almost "analog" feel to it. besides stuff like using correct directory in this case, it is closer to turning turning knobs than flipping switches.

It has been a year, have you tried ComfyUI? or maybe something else? I had started here when 1111 was seemingly the only answer. That said, I only came back to have sense of familiarity. That UI that comfy uses, is my favorite. The "wires" "connections" make everything so much more transparent. I know things are neatly nested in 1111 but not only is the alternative (if you keep it clean, mve stuff around, "cable management")way more intuitive/clear, but also more powerful.

I'm going to stay here for a bit in 1111 land, maybe give comfy a shot once I get some more of these basic ideas down here.