- This topic has 31 replies, 9 voices, and was last updated 2 years, 11 months ago by
.
-
Posts
-
Last few months there has been some interesting and very good information posted about SD here by members.
I thought It would be good, at least for now, to have a thread of its own.
This is good. I'd like to know more about it. I tried one once, midnight something, and I couldn't do a thing with it... I'd really love to be able to make pretty pictures with my renders like @Legolas18 and others are doing. I have no idea where to start though.
Legolas had a great link on how to install SD
Once you have done that You are basic-ly set to start, but you got to" feed" it something to work with
Loras and checkpoints are the things you need as far as I understand.Those can be found here: https://civitai.com/Now you got your AI something to work with
There are 2 fields:Prompts and negative prompts
Prompts is what you want in the pic, negative what you do not want
A complete how to do :Then there are 2 more things that I fiddled with
CFG Scale and Denoising strength
Those ( by the value you enter) decide how much influence the prompts have on an ImgageToImage generationI am a beginner to, but this is what I know, and got some nice results. But it is still a lot of trial and error for me.
But he, it is enough to get you started.
Good luck!
I have one here from July that morphed into a SD discussion thread. If anyone wants to continue the discussions feel free.
I thought that this might get a chuckle from those of us tinkering with AI via SD....
Along with the content I reckon a huge amount of AI art is also trash/trashy and/or pretentious. I would say the same for the Daz galley too. Not talking about image quality or amateurs just starting out and trying to do their best, that should be encouraged, but a general lack of effort or purpose I see so often.
All i can say is that I was a non believer, and had no interest in AI art. It took a comment made to a piece of art posted by Legsalot18 about image2image where Frank21 says "Very good demonstration of how to turn a Daz doll into a real looking human"
Thats when I started to look into Stable diffusion and how it works and now I'm loving every bit of it.Just spent all afternoon downloading pips and pythons and gods know what else, and I still have no Stable Diffusion and Automatic1111 or whatever!!
I'll keep trying, but I cannot find an actual download for them anywhere. There are pages on git hub but its only codes and things? Confusing to me.
Anyway, I appreciate all the info, and I'll keep trying! Eventually, MAYBE I will figure it out~
Checked out the model site too... of course it's mostly ALL women... very very few guys. Sigh.
All the best,
Niniane@ninianed, There's a goal for you. When you finally get it installed, create some male models. I've got a bunch of ideas for formal clothes and nice settings. All the anime, Asian characters and skimpware is a bit too much. It's kinda like Daz models.
And how many models of Taylor Swift and Emma Whats-er-name do we need?!@Ninianed - have a look through this. SD can be a bitch to install and get running properly.
Most of the models at Civitai are still for SD 1.5 but many are now for the newer SDXL version so decide which one you want to install. If you have limited disc space and a lower-end GPU I'd suggest 1.5. Loras for 1.5 are usually <100 MB and most are only a few meg, but the SDXL ones can be up to half a gB each. Good luck with the install 😊.Thanks guys, I still don't even know what a Lora is. 🙂 But I'll figure it out. I WILL. 🙂
@Frank22, that is exactly the page I used to download the git and python and some other things I can't remember. But I haven't found a SDXL install file? Where is it, how do I install? Why can't I find it? Ahh... I hate being so dumb. 🙂Most of what you see on the Civitai pages are Loras, which are typically characters or art styles.
Styles are fun to play around with.I'd like to remind folks of the AI discussion over at our Discord server where our lovely and talented host, AnjaBear, is always watching over our collective shoulders.
@Ethiopia - "And how many models of Taylor Swift and Emma Whats-er-name do we need?!" 😂
There are currently2830 Emmas at Civitai, but I'm sure there will be more, and I'd say most of them are not being used as she'd prefer 😱.LOL I guess this is true....
@Ninianed and anyone else interested....
I found this person to be very useful in sorting out the steps to take, including setting up a 'local install' on your own machine.
He has a lot of videos [as well as some bad jokes] and he is easy to follow.How to Install Stable Diffusion - automatic1111 Sebastian Kamph. He includes links to the required files in his text below the videos.
These 2 can get you going on the initial setup, and he has plenty more to help you develop.Good luck - hopefully you will have some nice AI art for the Xmas comp... 🙂
- You must be logged in to reply to this topic.
