[Stable Diffusion] Prompt Sharing and Learning Thread

[me3]

Nice to see highres fix still has it's extremely interesting memory usage...
1.9x uses much less than 2.0x and 1.8x uses more than 1.9...and 1.5x manages to OOM with the initial image creation for some reason...
Worryingly it seems that some of this poor memory management is spreading to the upscaling as well. At one point i had it OOM with trying to allocate an addition 6gb on a simple upscaling to just ˝1300x1300...


Still got a long way to go but it's getting there...

And the AI really doesn't like creating different "people", despite everything it keeps gradually moving things closer and closer until it at a bare minimum fits with the instructions...

 

[jberger990]

Has anyone done a good Adam Warren Lora? I checked CIVITAI and I couldn't find one

 

[me3]

Has anyone done a good Adam Warren Lora? I checked CIVITAI and I couldn't find one

If you mean the comic book writer/illustrator, it might already be there by default as a lot of "styles" are.

(rant not reflecting on quoted poster but meant as a general "comment")
That being said, imo, the more closely ppl try to "rip off" other ppls work or what they are "known" for, the more "misused" the AIs ability is, specially if they are trying to profit from it in any way as well. There's more than enough examples of ppl trying to claim "fan art" in very poor faith and "taste".
Same goes for the "misuse" of public personalities. If we think about it in a more old fashioned way, caricatures have been generally seen as a humorous and acceptable way to "misuse" likenesses so it's likely that the same kind of thing will be considered the same with AI generated scenes too.
It's pretty obvious that there's gonna start being lawsuit in the relatively near future over coping/stealing "art" or deformation for having used peoples likeness in situation they aren't known for. It's probably not gonna be a good idea to be on the receiving of those since there's limited existing laws.

So mix styles, create something new, that's kinda what creativity is about. If you put the likeness of a pornstar in those kinds of images you are far less likely to get in massive lawsuits than using someone not known for it, but it's so easy to create unique people by simply combining existing faces, nationalities, features etc that you shouldn't really have to.

Be creative, create, not steal

 

[Sharinel]

If you mean the comic book writer/illustrator, it might already be there by default as a lot of "styles" are.

(rant not reflecting on quoted poster but meant as a general "comment")
That being said, imo, the more closely ppl try to "rip off" other ppls work or what they are "known" for, the more "misused" the AIs ability is, specially if they are trying to profit from it in any way as well. There's more than enough examples of ppl trying to claim "fan art" in very poor faith and "taste".
Same goes for the "misuse" of public personalities. If we think about it in a more old fashioned way, caricatures have been generally seen as a humorous and acceptable way to "misuse" likenesses so it's likely that the same kind of thing will be considered the same with AI generated scenes too.
It's pretty obvious that there's gonna start being lawsuit in the relatively near future over coping/stealing "art" or deformation for having used peoples likeness in situation they aren't known for. It's probably not gonna be a good idea to be on the receiving of those since there's limited existing laws.

So mix styles, create something new, that's kinda what creativity is about. If you put the likeness of a pornstar in those kinds of images you are far less likely to get in massive lawsuits than using someone not known for it, but it's so easy to create unique people by simply combining existing faces, nationalities, features etc that you shouldn't really have to.

Be creative, create, not steal

But if you do want to steal

You must be registered to see the links

 

[devilkkw]

Also some other example in

You must be registered to see the links

. Based on sd1.4.

 

[Rovente95]

HI, can I ask you if SDXL is better of SD? If answer is yes, I can use LORA with XL?

 

[felldude]

HI, can I ask you if SDXL is better of SD? If answer is yes, I can use LORA with XL?

From a math technology standpoint it is superior. But it is 6GB and if you want to use the refiner that is another 5.9GB.

You can use lora's but if they where not trained on SDXL then they likely will not have a good effect.
If it is easy to train on it will probably surpass 1.5 (Unlike those 2.0 models)

Without out refiners in comfy UI I can do a native 1600x1200 image in 63 seconds (2.12 sec/it) with the SDXL model (With a 8GB RTX 3050)

 

[Rovente95]

From a math technology standpoint it is superior. But it is 6GB and if you want to use the refiner that is another 5.9GB.

You can use lora's but if they where not trained on SDXL then they likely will not have a good effect.
If it is easy to train on it will probably surpass 1.5 (Unlike those 2.0 models)

Without out refiners in comfy UI I can do a native 1600x1200 image in 63 seconds (2.12 sec/it) with the SDXL model (With a 8GB RTX 3050)
View attachment 2812497

what do you mean by train them and how do you do it? Furthermore, ultimately it weighs more but on the other hand the results are better if your system has the requirements?

 

[felldude]

what do you mean by train them and how do you do it? Furthermore, ultimately it weighs more but on the other hand the results are better if your system has the requirements?

Yes, for example their is a face set of data trained from Meta that is 100GB, (They tell everyone not to use it for facial recognition wink wink)
Most people can't do anything with it.

Training from scratch involves a 100k rig or more and usually takes days even with 80GB of video ram and 32 processors feeding that.

When us normal folk train a lora or checkpoint we are training with an existing set so in this case it would be SDXL, how it responds to training with 8bit optimizer or FP16, will determine the number of modified checkpoints or loras that support it

 

[Rovente95]

Yes, for example their is a face set of data trained from Meta that is 100GB, (They tell everyone not to use it for facial recognition wink wink)
Most people can't do anything with it.

Training from scratch involves a 100k rig or more and usually takes days even with 80GB of video ram and 32 processors feeding that.

When us normal folk train a lora or checkpoint we are training with an existing set so in this case it would be SDXL, how it responds to training with 8bit optimizer or FP16, will determine the number of modified checkpoints or loras that support it

forgive me I still don't understand what you mean by training... I only understood that Lora uses pre-existing models and XL uses Meta Human models

 

[felldude]

So the realistic vision checkpoint was trained against 1.5 SD for the purpose of drawing realistic humans. If the 1.5 SD checkpoint was only trained on landscapes to begin with, I could feed it 1,000,000 images of humans and it would still draw landscapes with human features.

For XLSD they claim the math is better

You must be registered to see the links

But it doesn't mean an exisiting model won't fit your needs.

Compare:

SDXL



in painting the SDXL image at .6 with realistic vision



Realistic vison need to in paint the image as it would have made a composite of two images at that high of a native resolution. (SD 1.5 was intended for 512x512 rather then 1024x1024)

 

[Synalon]

forgive me I still don't understand what you mean by training... I only understood that Lora uses pre-existing models and XL uses Meta Human models

To make a lora you "train" it by giving it a bunch of pictures of what you want it to be about from every angle. It then spends hours repeated looking at them and trying to recreate them, (thats the training part). After that if it worked well enough for your tastes its done.

Its actually far more complicated but thats the easiest way I can think of to explain it.

 

[TitaniumDickDiamondBalls]

Does anyone here have defacounter_mix10 model ? It used to be on Civitai, but was taken down from the site for some reason.

 

[Rovente95]

how can I paint a image?

 

[Synalon]

how can I paint a image?

You mean inpaint?

Put the image in img to img, and theres a tab below called inpaint. tick that and draw over what you want to change, select resize and refill and generate I guess.

 

[Rovente95]

Thanks everyone, Now I need to study "training" to XL

 

[devilkkw]

A little simple test with artist i like: Gris Grimly
Prompt: woman by Gris Grimly

Result with no negative:

Spoiler: no negative

You don't have permission to view the spoiler content. Log in or register now.

Result with

You must be registered to see the links

Spoiler: New-negative-v1.4

You don't have permission to view the spoiler content. Log in or register now.

And one with

You must be registered to see the links

Spoiler: Easy photorealism

You don't have permission to view the spoiler content. Log in or register now.

Image in png to check setting.

In general i don't like triggering artist, but for checking embedding or checkpoint i usually do simple test like it.
If i trigger an artist (with no negative) this do a 2d image, and getting something different like 3d or photo, i waste checkpoint.
This is a simple example, i'm currently testing checkpoint about all 2416 artist with this prompt.
Maybe i made a topic with all result if you interested.

You found all Artist on my

You must be registered to see the links

 

[me3]

For most people here SDXL most likely isn't gonna be "worth it", specially not if you're not already somewhat familiar with the basics.
Reasons being that very few things will work with it and the few things that does are still gonna be in a development and testing stage which creates a lot of oddities, bugs and other things that'll lead to results being different.
There won't be much "add ons" or embeddings of any kind compared to what you have with sd1.5 and as they show up they'll likely be less optimized, less accurate etc
A point that will apply more to newer users is that regardless of your system, you'll learn how prompts work, what doesn't etc much faster with sd1.5 than sdxl for the simple fact that XL requires a minimum size of 1024x1024, meaning you have to generate 4x the size each time you try to work out your prompt. Anything you learn about prompts will remain the same, basics are very unlikely to become that different and you'll have a much wider selection to play with, much more options in the results you can achieve and better fitted "tools".

 

[felldude]

I know I can't train on XLSD, maybe you could train with a 12GB card likely you need a 24GB.

And that is with optimizers, 40GB cards won't cut it even for LORA's if your trying to due full FP32 training.

Im guessing maybe 10% of the community will be releasing XLSD content.

I'm getting decent results using Realistic vision or another 1.5 checkpoint as a refiner. Average a minute for 1024x1024 keeping both checkpoints loaded in memory with Ergan4x+ as the upscaler.

 

[Mr-Fox]

Yes, for example their is a face set of data trained from Meta that is 100GB, (They tell everyone not to use it for facial recognition wink wink)
Most people can't do anything with it.

Training from scratch involves a 100k rig or more and usually takes days even with 80GB of video ram and 32 processors feeding that.

When us normal folk train a lora or checkpoint we are training with an existing set so in this case it would be SDXL, how it responds to training with 8bit optimizer or FP16, will determine the number of modified checkpoints or loras that support it

That sounds very interesting. Do you have any links to what you are talking about so I can do some research?