- This topic has 39 replies, 15 voices, and was last updated 2 years, 11 months ago by
.
-
Posts
-
* as the has now moved into essentially a SD discussion I've changed the thread title
Apologies for being MIA for the past month. Part of it is being unexpectedly called away OS (Tokyo) for 3 weeks to see relatives. Missed the art competition 😒. And the other part is my preferred artistic style, always has been, photorealism. The sort of realism I can't achieve in DS and that has been superseded by SD.
Before I go on I'm well aware that there is a certain club mentality regarding software and hardware and artistic styles that can lead to bad feelings when one moves from one to another and I hope we can rise above that. I've made too many good friends here to walk away and I'll continue to be a regular supporter and contributer to ZGFX.
As for SD, it's no lazy alternative for the artless masses. Getting two people in a scene is very difficult and having them interact with each other innumerably harder than DS. While I could make half a dozen or more DS renders per session with SD I'd be lucky to get two I'm happy with per day.
Well they are certainly beautiful renders. 🙂 I can't get anything but a big mess when I tried it, so I just gave up! I Guess it's not as easy as 123...
Anyway, make sure you drop in on us once in a while so we won't think you went down with the Titanic. 🙂
All the best,
NinianeMy intention is to remain an active ZGFX participant so you won't get rid of me that easily 😛.
I'm glad you're back. I didn't like the competition so much. Mei lin is not really my cup of tea. But I think you did a good job in the SD. I gave up on that a long time ago, actually. It's really not my thing.
And it's true. Never have a drink and eating sweets with Russians on the Siberian Express. That's what I did.Nice to see you back, Frank.
Those are looking really good.
Are they G8F ?
BTW I do hope you will continue with your fine prop conversions to DS.
Best regards,
Eel 🙂
@frank22 Welcome back I guess then 😀
@eelgoo AI models 😉
You can get pretty good models and guides here for Stable Diffusion (SD) -> https://civitai.com/OK Thanks for the info.
I think I would like to retain more direct control of what I am doing than trust to AI.
:0/
The text-to-image generation is still a bit too random or at least has a large variance for my liking. But the image to image generation is amazing, with a little work it can turn your DAZ renders into "something more" 😉 With the right models you can mod the shit out of real photos too 😀
Nice work @frank22. What checkpoint are you using? I've flip-flopped between a couple but, for now, I've settled on cyberrealistic v3.2.
I too am heading down the SD tunnel. There's soooo much information available that I feel I'm making much more headway that fighting with Studio and it's lack of meaningful documentation and proprietary everything. I may wind up using Studio for nothing more than providing poses.
Check out the AI thread in our Discord server.Welcome back and congratulation for the switch! It's never easy to leave a comfort zone and discover a completely new field of expertise in which… you have none when you start 🙂
I've been flamed on some Discord server by people like the ones you mentioned (and that you probably won't find in our community): people saying that AI Art is not art and that clicking on a button is not creation. The first time I started digging in loras, prompt "engineering", clip skipping and tried to understand how all clicked-in together, I was as lost as the first time I put my hands under the hood of the Iray rendered… confidence requires practice… as does art. Limiting yourself to a few models and loras instead of downloading all the new ones without discernment is as hard as fighting the appeal of all the 3D assets (of varying qualities) that appear on every online shop.
If you put aside the people who click randomly to get some "more than average" NSFW as fast as possible, I think a new form of art is emerging. The technique to be learned is not holding a brush or applying colors by hand, it's prompt crafting. A single word change, a comma added, a change of order in the sequence of keywords, and the result can be entirely different. Knowing your models and the way they react to some words, positive or negative, is also part of the learning process.
I'm not too afraid though because the same people complaining about AI Art are the same kind that bellowed when Photoshop (and alike) and digital art emerged… and we don't hear them complaining anymore. They feel "threatened" in some way, thinking AI will take over when, in reality, all forms of art are complementary. As for letting DS go… I would advise otherwise. Of course, you'll probably put a stop (or a good braking) to asset collecting, but you have probably already discovered that (even if it's no allowed by the new DAZ TOS), with controlnet, you can use DAZ to pose a complex scene and use it in SD very effectively. As you mentioned, having two characters (or more) interact the way you want by the simple "strength of your prompt" is not an easy task yet. But the technology is evolving so fast, who knows what we will be able to do in a few months, even weeks considering the pace.
To conclude because I feel I've already written to much, and like @ethiopia said, check out our AI thread on the Discord. It's not very active at the moment but only awaits people to grow 🙂
I'll also tag @randykoala who is even more into SD than me and who is producing really neat images and testing all the fancy extensions that appear. Try the Photon Checkpoint if you haven't yet. More natural prompt and very low settings that render fast. Finally, never let anyone tell you that this is not art 🙂A random generated woman, but with 100% my wife's face on her. The extension for that is called roop.
Have a great day and don't become a stranger 😉
Math
Hey Math, your wife's a real sweetheart. Her expression is hard to make out though. It seems to say,
"Get that camera out of my face before I hurt you!"@ethiopia - she doesn't like the camera very much, you are right. The same at home maybe ? 🙂
SD is the acronym for Stable Diffusion, an AI based image generator.
You might know Midjourney that is not free anymore and only runs online? SD can be installed on your computer (provided you have a decent enough graphic card) and, after downloading models (basically huge files that are the result of the training of the AI after being fed thousands of images), you can "prompt" (by literally writing the description of what you want the AI to generate) SD to produce images.
SD has no interface per say so people install different open sources interfaces among which the best know are Automatic1111, Vlad Diffusion or Comfy UI (just to cite those).
Everything is free, well documented on the web and pretty fun and packages exists that include the SD engine and an interface to use it (via a browser).
If you follow this link, go to the chapter "Automatic Installation on Windows" on the page and follow scrupulously the 4 steps described, you'll be up and running in an hour or so.
After that, you can head to Civitai where you will find loads and loads of models (checkpoints, Loras, textual inversions...) and to get a grip on all that you can go to those Youtube channels :
- Olivio Sarikas
- Sebastian Kamph
- AitrepreneurBut remember, you need a good graphic card (4 GB of VRAM minimum), preferably with an NVidia GPU, and a good internet connection because we are talking Gb and GB of downloads 🙂
Cheers
Math
- You must be logged in to reply to this topic.
