It’s true, lol. I forgot to add some wildcards for the breasts.

😂

Oh man, you’re a great source of knowledge! I was wondering why I wasn’t getting more diversity! Also, I didn’t know about wild card files! I’ve got some homework to do! Thanks!

Oh sweet. Thanks, I’ll check it out…

Humm, good point. Thanks!

It surprising how hard it can be to get good cameltoes or pokies…

Thanks! I really like this set…

Yeppers!

Haha, yes, I’m copying them lol - but the interesting thing is that a negative prompt has the potential to remove things other than the thing the prompt implies. For example, if you add ‘ginger’ to your negative prompt you’ll get different results than if you add ‘red head’. So, in short, if I find a prompt I want to copy, I copy both the prompt and negative prompt to increase the likelihood of getting results similar to the image I’m trying to replicate.

The thing that’s tempting me is the possibility of being able to combine an XL Lightning model with an animation module (something I can’t seem to be able to do in Automatic). Just haven’t pulled the trigger yet…

Edit: I just realized who I’m talking to! Of course you’re familiar with Comfy UI - lol.

Interesting, I’m guessing you’d probably want to use Comfy UI over Automatic 11111 for a flow like that, but I’d love to hear about it if you end up doing something along those lines!

If you click on an image it should open to a webpage with the prompt and generation parameters including the model. But honestly, only about a third of the ones I generated had good hands lol.

I don’t recall having to manually turn that feature on. (I’m on Windows 11). If I recall, Windows will only allow a process using VRAM to overflow into system RAM up to half of your total system RAM. So, if your system has 32GB, Windows will only let you use 16GB. If it’s crashing, you can try generating your images and then upscaling them afterwards in a separate process…

Here’s more details on my workflow. https://lemm.ee/post/28542626

Absolutely, I’ll post a screenshot of my settings in my albums going forward. Thanks for the feedback!

You’re most welcome!

SDXL moves from a base resolution of 512x512 to 1024x1024 along with better prompt recognition, so your prompting will likely need to evolve. However, there is a new type of SDXL model, ‘lightning’ that generates images in 2-8 iterations as opposed to 16-60. Prompts are also different for these types of models. I suggest the SDXL Lightning version of Juggernaut. The only time you might be interested in going back to older models would be compatibility with Loras or animation modules.

Note that on lightning models you’ll need to lower your CFG Scale to around 1.5 - 2.5

https://civitai.com/models/133005/juggernaut-xl

Wow, great work! I’ve been dragging my feet on Comfy UI. It looks like it’s worth it!

Glad you like it! For some reason when I added ‘ponytail’ to the prompt it changed the woman significantly, so I opted to leave it out.

Done! https://lemm.ee/post/28359066