- Mar 18, 2019
- 1,251
- 1,764
- Jan 24, 2020
- 1,401
- 3,808
Ikr. I only posted it to show people that it's not so hard to generate an adult woman with low body fat.. Nothing feminine about it imo and a complete turn off.
That's just..anorexic
As for your negative prompts you posted, I use most of them, a couple I didn't even know as words. But no positive prompts to enforce it.
- Mar 18, 2019
- 1,251
- 1,764
Ok man, that's not a review...
It's just because it hurts my gut to see a potential beautiful girl like that. It's a cold react!
- Jan 24, 2020
- 1,401
- 3,808
After a bit of tuning of prompt and settings.
Are we getting warmer?
Are we getting warmer?
I've been trying on my end too. Removed all the bloody references that could even insinuate a minor, then thought the model might be the issue, changed to Clarity. I either don't like the model at all, or I'm doing something very wrong and get a very 3d like appearance, too plastic-y, earlier model looked a lot more realistic.After a bit of tuning of prompt and settings.
Are we getting warmer?
As for the issue at hand.. You tell me, but I don't think it's doing much.
This is the render without any post processing, so all prompts and setting should be visible, if you (or anyone else) have any ideas. I want to be done with it and create the character I have in mind. And no, it's not a minor.
Though I have to admit, Clarity manages the [Hawaiian|Scandinavian] much better, as far as face is concerned. Too bad it falls behind on realism.
- Jan 24, 2020
- 1,401
- 3,808
Try different samplers (sampling method). Heun is one I'm very fond of, DPM++ SDE is great also, try out the different ones that has ++ and karras in the name and DDIM. PLMS and LMS etc has never got me any good results. Also try different amount of steps and cfg. I'm running a Checkpoint comparison with the last prompt and seed that I posted. Coming very soon..I've been trying on my end too. Removed all the bloody references that could even insinuate a minor, then thought the model might be the issue, changed to Clarity. I either don't like the model at all, or I'm doing something very wrong and get a very 3d like appearance, too plastic-y, earlier model looked a lot more realistic.
As for the issue at hand.. You tell me, but I don't think it's doing much.
This is the render without any post processing, so all prompts and setting should be visible, if you (or anyone else) have any ideas. I want to be done with it and create the character I have in mind. And no, it's not a minor.
View attachment 2438840
Though I have to admit, Clarity manages the [Hawaiian|Scandinavian] much better, as far as face is concerned. Too bad it falls behind on realism.
- Jan 24, 2020
- 1,401
- 3,808
I checked the PNG Info. You can't use tags like "e-girl" or anything other with the word "girl". That's the main culprit.I've been trying on my end too. Removed all the bloody references that could even insinuate a minor, then thought the model might be the issue, changed to Clarity. I either don't like the model at all, or I'm doing something very wrong and get a very 3d like appearance, too plastic-y, earlier model looked a lot more realistic.
As for the issue at hand.. You tell me, but I don't think it's doing much.
This is the render without any post processing, so all prompts and setting should be visible, if you (or anyone else) have any ideas. I want to be done with it and create the character I have in mind. And no, it's not a minor.
View attachment 2438840
Though I have to admit, Clarity manages the [Hawaiian|Scandinavian] much better, as far as face is concerned. Too bad it falls behind on realism.
Use female (I noticed that you did) or woman. You can specify age or use "adult". About the plastic look, try to switch Face Restoration to codeformer. Try both and see wich is best. Though it might depend on wich checkpoint you are using.
Last edited:
I thought it might be that. Would be the next thing to try even if I lose a bit of the concept. But that's a problem for tomorrow. Thanks for the help.
- Jan 24, 2020
- 1,401
- 3,808
Just googled E-girl because I didn't really know the meaning.
What is an e girl aesthetic?
The e-girl aesthetic is typically characterized by winged eyeliner, heavy eye makeup, a childlike quality, and often, colored hair. "Some characteristics I would say are definitely things like darker hair or hair with lots of color, dark makeup whether a boy or a girl and carefree hair!" says @peachyjaii.20 Feb 2020
Explaines a lot imo. Notice the phrase "childlike quality".
What is an e girl aesthetic?
The e-girl aesthetic is typically characterized by winged eyeliner, heavy eye makeup, a childlike quality, and often, colored hair. "Some characteristics I would say are definitely things like darker hair or hair with lots of color, dark makeup whether a boy or a girl and carefree hair!" says @peachyjaii.20 Feb 2020
Explaines a lot imo. Notice the phrase "childlike quality".
Couldn't let it go so I removed those references and made another one. You were right. As for that definition, I didn't know that, I had twitch streamers in mind when I put it.
That one looks closer to late 20s to me, but I cba anymore today. Might be it's totally fine and my eyes are fucked from all the renders today, lol.
Nipples fixing any prompt spamming until I get it right tomorrow. I need to sleep.
That one looks closer to late 20s to me, but I cba anymore today. Might be it's totally fine and my eyes are fucked from all the renders today, lol.
Nipples fixing any prompt spamming until I get it right tomorrow. I need to sleep.
- Aug 24, 2022
- 313
- 744
For those who are using control net and want a good way to control lighting:
(It's a short video and super simple if you already have control net)
Also if you ever wondered why AI images look sort of washed out and are lacking good contrast, this will soon be a thing of the past, see here:
(It's a short video and super simple if you already have control net)
You must be registered to see the links
Also if you ever wondered why AI images look sort of washed out and are lacking good contrast, this will soon be a thing of the past, see here:
You must be registered to see the links
- Aug 24, 2022
- 313
- 744
Just playing around with Control Nets..
Seeing What poses I can make with as little effort as possible (Because I am lazy)
I count five fingers but I think the hands need a little clean up still.. I haven't fed a depth map in yet though, this is only posing.
Depth Mapping hands seems to fix them 100%.
Seeing What poses I can make with as little effort as possible (Because I am lazy)
I count five fingers but I think the hands need a little clean up still.. I haven't fed a depth map in yet though, this is only posing.
Depth Mapping hands seems to fix them 100%.
- Nov 16, 2017
- 217
- 3,124
It feels rather strange to combine Stable Diffusion with ChatGPT...
- Aug 10, 2016
- 69
- 146
Thanks! Using this technique I was able to change the time of day in a background image way better than I thought possible
- Aug 24, 2022
- 313
- 744
time to do .. Jack-O Challenge as an experiment.
Using Open Pose only
Almost identical prompt to the previous image.. just posed differently (might be different seed too)
Hands look like a jumbled mess..
Open Pose + Depth Image for Hands
(But different seed & prompt.. whoops)
Conclusion
There are some things I like about the first image more.. I changed the seed and prompt a little.. (durp)
The biggest takeaway is that the hands don't look like cthulu-like tendrils of flesh anymore.
(Also the ability to bend almost any drawing style / checkpoint to match the exact pose you want using a Rig is super powerful for context-aware images, with lighting & normal maps > Control Net has turned Stable Diffusion into Daz basically. lol
Using Open Pose only
Almost identical prompt to the previous image.. just posed differently (might be different seed too)
Hands look like a jumbled mess..
Open Pose + Depth Image for Hands
(But different seed & prompt.. whoops)
Conclusion
There are some things I like about the first image more.. I changed the seed and prompt a little.. (durp)
The biggest takeaway is that the hands don't look like cthulu-like tendrils of flesh anymore.
(Also the ability to bend almost any drawing style / checkpoint to match the exact pose you want using a Rig is super powerful for context-aware images, with lighting & normal maps > Control Net has turned Stable Diffusion into Daz basically. lol
