- Nov 17, 2017
- 28
- 74
This is a first attempt with my complete lack of knowledge of AI art.
- Nov 17, 2017
- 28
- 74
One more attempt. This is probably what happened after the last image.
- Nov 17, 2017
- 28
- 74
here's two more that looked ok, still trying to figure the prompts out for generating.
What tools, model, LORA, promt (both positive and negative) did you use? If you use my LORA I can give some tips.
- Nov 17, 2017
- 28
- 74
i used the civitai link and just created using that, i'm a complete noob with the technical side. prompts were mostly trying to best describe the scene, using as many phrases to pinpoint what i wanted. negative prompts was things like black and white, sketch, and rough edges, or more that two people, because it kept giving me very rough looking art at first.
Then you should read my tips on Civitai about LORA. It has advices about positive and negative promt, as well as tags for different Amelia clothes from the game. You can also check my example arts on Civitai. Each art has metadata with info about promt, settings, which were used to generate it. Don't forget to use Illustrious base model.
- Nov 17, 2017
- 28
- 74
I use ComfyUI. Still hard to distinguish two characters so I usually made the man out of frame with just parts of his body visible. Any tips?
I am using Webui.
For two characters I suggest to look onto example arts at LORA page. It has metadata with all key settings and promt which I use. Remember that I also use 4x UltraSharp and Adetailer to fix face generation artifacts and increace overall quality of picture.
- Jan 11, 2022
- 309
- 1,890
Best way to get two distinct characters in a shot is to use some kind of regional prompts. In A1111 they call it Regional Prompting, Forge calls it Attention Couple. There's a few custom nodes in Comfy to do it, usually called something along the lines of attention mask or regional mask. Like everything in Comfy it gets complicated quickly.
Once you get it figured out it's a game changer though. Probably my favorite tool for AI CG.
- Jan 11, 2022
- 309
- 1,890
You must be registered to see the links
That's an example. There's so many custom notes out there for ComfyUI there might be a better one by now. But it shows the gist. You mask different parts of the image and say "redhead over here" and "blonde over here", etc.
You can also just make a regular image with two people, then use inpainting to retouch a portion of it. So your initial image gets two redheads, but you run it through an img2img workflow and turn one of the redheads into a blonde.
Hope to see more art in this thread of our favorite girl taking her trainers exercise rountinely.
- Nov 17, 2017
- 28
- 74
- Nov 17, 2017
- 28
- 74
It was a busy day at work for Ammie.
Then her hubby texted to chat for a bit.
She was told to use her rather LARGE talents.
All in all, a hard day's work for Ammie.
Then her hubby texted to chat for a bit.
She was told to use her rather LARGE talents.
All in all, a hard day's work for Ammie.
Looks like she'll be used quite a bit more and might be staying a bit late.
- Nov 17, 2017
- 28
- 74
thanks, you should have seen some of the modern art masterpieces the AI chucked out alongside these... stuff of nightmares.