- Dec 4, 2022
- 6,533
- 58,356
- May 7, 2019
- 666
- 5,217
- Sep 20, 2021
- 918
- 4,478
- Oct 1, 2022
- 2,205
- 11,924
- Dec 18, 2022
- 44
- 339
- Oct 1, 2022
- 2,205
- 11,924
- Dec 4, 2022
- 6,533
- 58,356
- Oct 1, 2022
- 2,205
- 11,924
- Oct 1, 2022
- 2,205
- 11,924
- Sep 20, 2021
- 918
- 4,478
- May 7, 2019
- 666
- 5,217
- Dec 4, 2022
- 6,533
- 58,356
- Jan 31, 2024
- 1,199
- 3,009
- Oct 20, 2021
- 220
- 628
- Mar 18, 2020
- 711
- 3,628
Was testing different ethnicities, faces tend to end up kinda samey. Will have to get some loras to achieve more accurate/varied faces. Still nice though.
Latina(got a few pretty good ones)
Ebony
Asian (does better but have to force adetailer to adhere using another prompt)
Indian(looks more ebony than indian)
Latina(got a few pretty good ones)
Ebony
Asian (does better but have to force adetailer to adhere using another prompt)
Indian(looks more ebony than indian)
- Jan 31, 2024
- 1,199
- 3,009
- Mar 18, 2020
- 711
- 3,628
It's called an innie pussy, the clit is not necessarily visible.cute faces, but the pussies with missing clits looks a bit weird
Edit: A quick google search will provide many real world examples.
Last edited:
- Oct 3, 2021
- 1,518
- 10,190
- May 14, 2022
- 968
- 5,748
Most of this stuff is epicrealism, these ones didn't use a lora. These are older too, SD 1.5
Even though pony and XL and flux and stuff are all newer and shinier and in fact (let's be real) better, they do struggle with ethnicities and distinct faces out of the box. LoRas and regionprompting and a lot of prompt trial and error can do it of course but it is one thing that's bothered me for a while.
Opposite direction: low tech low spec low fidelity low detail heh heh
- Dec 18, 2022
- 44
- 339
