SD1.5 try size: 640 x 640 or 512 x 768
SDXL Try size: 1024 x 1024 or 768 x1024
Weights: 0.3~1
Clip skip: 1~2
Training material from bing
Pictures from CalicoMix and AnimeXL-xuebiMIX and 古風 GuFengXL_Turbo_Mix and VXP AnimaPonyXL (Turbo) and Real Lightning MIX - 4 Steps
Note: If girl disappears, please add more description in SD1.5, and sometimes girl has machines in her hands. SDXL lora had slightly overfitting (numRepeats:16), not suitable for realistic model; Pony lora versatility is higher (numRepeats:8) , more suitable for realism model. If you use ADetailer, please to close "ADetailer VAE". and "Inpaint mask blur" should not be lower than 8.
Description
FAQ
Comments (21)
ultra basic prompt....
"pedestrian-crowd", "Pedestrian-overload", "pedestrian-crowd-overload", "city-life-overload" and a lot more simöle basic prompts that will generate the same....
"Countless" is quite close to "Overloaded", and the results are not much different.
The effect is not obvious in the SDXL realistic model, but the effect is obvious in SD1.5 checkpoint, especially semi-realistic.
In fact, I don’t need too many pedestrians, the focus is on traffic issues.
if it's just traffic, then check my uploads and those prompts....
@Stapfschuh02 Sorry...I don't really need these prompt. I have a group of friends who specialized research in prompt, and I also specialize in this area. I can easily draw a lot of things without relying on lora, such as drawing countless garbage mountains and garbage seas in any model, because that is the feature of SD1.5.
Lora is to enhance accuracy, as a helper, it is not the enemy of prompt or checkpoint.
@ghostpaint prompting such things is a super simple task if you know how to prompt it. you are right with that.
@Stapfschuh02 I have a lot of similar works, but that's not the point of lora testing.
@Stapfschuh02 Sometimes it’s hard to control your desire for expression, right?
@Stapfschuh02 It’s not because it’s difficult, so I training Lora.
@ghostpaint xD, i totally agree
@Stapfschuh02 I admit that you are an expert, but there are many things that I have not shared, doesn't mean I don't understand.
@ghostpaint i'm absolutely not an expert. it's just that any answers instantly pop up inside my head as if it would be something normal or common...
you can see those results here JadeEpicPortraitWorldXL_V1.1 - v1.0 | Stable Diffusion Checkpoint | Civitai
or here Civitai | Share your models
or here [Sammelthread] - stapfschuh's new home for sharing prompts | WoH Board: Anime & Hentai Forum (world-of-hentai.to)
@Stapfschuh02 I am too lazy to write so much. If I want to write, it will be in Chinese...For example, how creating enough headroom.
@ghostpaint headroom? how is that meant? i know the meaning of this word, but not the context right now
@Stapfschuh02 If you type "headroom", Will there be enough space above the girl's head? I mean I know your undiscovered blind spots.
@ghostpaint it will most likely generate some kind of room. if you need that headroom, you can prompt "gigantism" in negative and "outside" in positive
@Stapfschuh02 I have a way around it but you can't. The background must be painted partially first, the girl will appear later.
@ghostpaint can you explain this a bit more? right now, i have no time anymore as i must go to work. will be back in 12 hours....
@Stapfschuh02 I'll show you sample picture. Normally, I won't put pictures that are not related to lora here, include other people. This is not the place to discuss prompting, especially if it has nothing to do with lora. If you think this is boring, you don’t need to force yourself to solve it, or wait for the answer to appear in your head. I don’t want to explain, I didn’t use any tools or lora, only prompt.