Still adjusting settings. Got a workflow setup for rapid generation so expect a fast turnaround.
You supply the faces.
I supply the facials.

>>936615065

motionmuse.ai/r/hp9b2bt7

>>936615122
That was my work from last night! Glad to see you're a fan

>>936615237

>>936615191
Yeah, that was a good time.

Grabbed a few images from other threads while I wait for requests to come in.

>>936615543

>>936615587
thanks

>>936615558

>>936615579

>>936615602

>>936615613

>>936615643

>>936615659

>>936615674

>>936615711
thanks again

>>936615749

>>936615821

>>936615823
Plz do my cousin

>>936615863

>>936615823

>>936615913

>>936615872

>>936615878

>>936615897

>>936615755
So sexy

>>936615907
Bro I don't care if it's your cousin. This ain't a themed thread. I'm honestly barely even looking at these. I just check to make sure it'll give a half decent result then I queue it to run once and post whatever comes out.

>>936615922

>>936615934

:)

>>936615955

>>936615988

>>936616032

>>936616063

>>936616107

>>936616120

>>936616128

>>936616157
Hmm... Not sure if it was the angle or the vape but that didn't work right

>>936616145

>>936616154

>>936616185

>>936616188
try this one instead?

>>936616193

>>936616232
Oh yeah, before I even queue it up I can tell that'll work a lot better

>>936616209

>>936615020 (OP)

>>936616287
Honestly surprised this worked as well as it did with the eyeshadow. Currently using two different cumshot loras in sequence to get this result. Think I might add a third to see if I can get streaking mascara too. Is everyone ok with me taking a break so I can add to my workflow real quick?

>>936616225

>>936616232

>>936616252
Good job with the thread mate

>>936616367
much better, thanks wiz

>>936616292
I honestly thought that was what's-her-name from the fallout show at first. Now I'm surprised no one has posted her yet.

>>936616293

>>936616300

Please

>>936616304

>>936616313
>>936616304
Excellent choice for picture btw. I honestly laughed a bit

>>936616323

>>936616325

>>936616481
Thanks, looks like xochitl's big lips give it trouble

>>936616340

>>936616546
Yeah, if I were to take a little more time and a second pass I would have lowered the denoise by a couple percent to keep her features more intact. It would have been less cum though and it's already only set to 0.18

>>936616348
Whichever anon is posting the ones with the very appropriate text: I see you, I appreciate you

>>936616375

>>936616413

>>936616474

Tip for OP, you can add wetspots to their shirts for extra detail in authentic looking way.

>>936616510

Put some cum on her little pussy please?

>>936616687
Hmmmm... I'm doing inpainting or adding a depth mask for this, it's literally just img2img with a few specialized loras. I built this workflow for speed. I already have a lora for transparent clothes but I've only been able to get it to work on higher denoise strengths with inpainting. I might be able to find or build a lora for wet clothes though and integrate that. I think my next thread is going to be "Messy Facials" and will include not just the ones I'm relying on now but adding in one for streaking mascara and wet clothes

>>936616532

>>936616561
Full body doesn't seem to really work that well with this workflow

>>936616567

>>936616613

>>936616634

Mom and daughter?

>>936616718
I'll be real with you, I'm not sure this is gonna work. I tried to get it to do cum on toes for about 2 hours the other night and couldn't get the ai to behave. I'll give it a shot but no promises

>>936616650
Nice

>>936616682

>>936616882
Ty!

>>936616706
Ahh, it fucked up the text. Sorry

>>936616718
Worked better than I expected honestly

>>936616926
It’s good thank you

YO! We're starting to run out of image limit. I need you all to start deleting your original request after you get the response and slow down so we can fit everyone

>>936616723

>>936616725

>>936616759

>>936616796

>>936616817

Please cover Nami

Please

>>936616827

>>936616858

>>936617128

>>936616896

>>936616902

>>936617196
Hmm... I think adding a depth mask would help, but I'm guessing it would add at least 10-15 seconds to a 45 second gen... I'll do some testing before I run the next thread

>>936616965

>>936616968

>>936617058

>>936617080

>>936617083

>>936617085

>>936617109
This would have worked better with a closer face shot. Also she had the eyeblack on

>>936617196
>>936617246
Thanks mate!

>>936616920
My bad, missed this one

>>936617163

>>936617425
Only have room for a few left in the thread. Did I miss any?

>>936617481

>>936617532

>>936617528
Ty! To close it out?

We've got room for one more.

>>936617654
Nicely done

>>936617704
Ty for all your hard work

>>936615020 (OP)
also what are the technical limitations of this tech?
eg how much data do you need to start making images? how much processing power does it take? how quickly does it iterate? what's the resolution? would it be best to render a whole scene then process the facial data one by one afterwards

OK. Some notes for next time:
Fix1: adjust my auto-resizer group to see if I can eliminate that black border that shows up sometimees.
Fix 2: Add a couple extra loras for streaking makeup and wet clothes for even messier facials
Fix 3: Add a depth mask for better face detection
Fix 4: Make a dual workflow so I can switch between normal facials and messy

>>936617899
I built my workflow to automatically downscale almost any size of image to something that my SDXL model can handle. It then upscales it back to the original resolution afterwards.
The upsides of this mean that I can quickly process almost any size of image and output it with less warping than I would if I tried to adjust that image at its native resolution.
The downside is that the bigger the disparity between the original resolution and 1024x1024 the more it'll mess up during the upscaling.
Still though, as you can see it handles results pretty well.
For processing power I'm literally running off an old laptop with a 6Gb Vram RTX3050 and 16gb Ram and I still get 45 second gens