cDupin
Member
Posts: 476
|
Hi everyone, as many of you might know I started doing AI pictures about a month ago. I recommend and encourage everyone to try it out for themselves as it is surprisingly easy to get started, once you know how it's done. Since a few people have shown a tiny bit of interested, I thought it might be a good idea to start a general AI picture thread. My idea is that we can ask questions here, give advice to each other and share images we have created. To get things going I wanted to show you an assortment of images that I created before I started doing the comic series. To figure out how everything works. I thought you might like some of these. Anyway, I hope this is interesting to you guys, if not I'll just use this to upload a few pictures from time to time. So here we go, the first pictures I did (as you can see, they are all the same motive)
|
cDupin
Member
Posts: 476
|
After that worked I tried a few different settings and characters
|
cDupin
Member
Posts: 476
|
And a few different "activities"
|
cDupin
Member
Posts: 476
|
As you can tell it's quite a different style compared to the Scarlett series. Which is something I absolute about AI art. You can try and do soo many different styles and it's always exciting how it will turn out. Anyway, I don't want to spam this forum. Let me know if you are interested, I am happy to explain things and/or help people to set this up. Or if you just want to enjoy more images I am happy to upload more and show you the possibilities. Cheers guys
|
creempieluvr
Member
Posts: 74
|
What site do you use. I would love to try this.
|
cDupin
Member
Posts: 476
|
creempieluvr: What site do you use. I would love to try this. I explained it in the other threat but I am happy to wirte it down here as well - maybe I should have in the first place. I run stable diffusion locally on my PC. This has several advantages compared to a website: 1) No restrictions in regards to NSFW Content. You can prompt whatever you like 2) No costs (other than electricity, obviously). I don't want to spend money on doing this, no thank you 3) More control. A lot of it is still very much luck based (for me at least) but once you understand a bit of it and do a little research you have a lot of control in regards of what style you want to use and how everything should look - most NSFW AI images sites out there are utter garbage imho. At least if you want to do "special interest" stuff and not just an avatar of a hot AI chick How is it all done? I just asked perplexity.ai (my favourite chatbot, I am sure others can do the same) how to do NSFW AI pictures locally and it walked me through the whole process. You just have to install stable diffusion, download a few models you like and off you go. I had ZERO knowledge about this but if you have a few hours of uninterrupted time it's easily done in an afternoon.
|
MrBigCuckold
Admin
Posts: 5964
|
|
nightbirds_2
Member
Posts: 162
|
#8 · Edited by: nightbirds_2
|
 |
cDupin: Let me know if you are interested, I am happy to explain things and/or help people to set this up. Or if you just want to enjoy more images I am happy to upload more and show you the possibilities. thank you for starting this. yes please explain things if you can. i would like to learn more but dont know where to start. my other question is how to make the pictures look more realistic instead of looking too cartoony or illustrated. thank you.
|
cDupin
Member
Posts: 476
|
#9 · Edited by: cDupin
|
 |
nightbirds_2: i would like to learn more but dont know where to start. I recommend talking to a chatbot like perplexity.ai or chatgpt as you can ask them to help you fix errors you might come across when trying to set everything up. You need to install stable diffusion - this is the software that makes the images. Once it's running you need to download models (called checkpoints) form which the software takes the input to process your prompt. To quote perplexity: You can download the AUTOMATIC1111 Stable Diffusion Web UI from its official GitHub repository:GitHub Repository: The hyperlink is visible to registered members only!This is the most up-to-date and trusted source for the software. Installation Steps (Windows Example) Install Prerequisites: Python 3.10.6 (not newer) (edit: The hyperlink is visible to registered members only! Git (edit: The hyperlink is visible to registered members only!Download the Web UI: Open a terminal or command prompt and run: git clone The hyperlink is visible to registered members only!Download a Model Checkpoint: Download a Stable Diffusion model (e.g., v1.5) from trusted sources like Hugging Face or CivitAI. Place the .ckpt or .safetensors file into the models/Stable-diffusion folder inside your stable-diffusion-webui directory1 Run the Web UI: Double-click webui-user.bat in the stable-diffusion-webui folder. The interface will open in your browser at The hyperlink is visible to registered members only!Here are a few screenshots to show you what it looks like:
|
cDupin
Member
Posts: 476
|
nightbirds_2: my other question is how to make the pictures look more realistic instead of looking too cartoony or illustrated This is primarily done through the checkpoints (and loras) you use. Checkpoints are sort of the "main model" stable diffusion uses to "draw inspiration" from. To show you what I mean, I have created a few images with the exact same setting (even the same seed) but using different checkpoints. As you can see, the results are vastly different. I have only used checkpoints I have already installed, which is why they are more "comicy" as this is the look I prefer. I haven't looked into photorealistic stuff which is obviously something you can try and do as well. I recommend The hyperlink is visible to registered members only! there are loads of free models and loras (you might have to open an account and go to the settings to see the nsfw stuff). It is also great to get inspiration and see how other people have created their images - this is for the next lesson though Let me know if you've been successful in setting everything up.
|
cDupin
Member
Posts: 476
|
One more thing: as you can see in my example pictures, a lot of it isn't quite right. So if you wanted to get a better picture you could try improving your prompts (eg. I would usually do more in the negative prompts) or run several iterations with different seed until you get it right. There are also a few things you could do to "fix things in post" but this, too, is for another lesson. In the mean time, here are a few more of my first try to make class less boring 
|
cDupin
Member
Posts: 476
|
|
bcs
Member
Posts: 3552
|
These are fantastic!
|
fred01
Member
Posts: 3404
|
cDupinMerci pour ce debut de lecon, tres interessant et bien envie de me lancer aussi pour m amuser! d'autant plus que tes resultats *** vraiment tres blufant   Peut tu aussi creer d'apres des per***nages existant? fred
|
cDupin
Member
Posts: 476
|
fred01: Peut tu aussi creer d'apres des per***nages existant? Je ne parle pas le francais! Je l'avais à l'école, mais je ne me souviens presque de rien. I hope google translate did a good job so I can answer your question: yes, you can do it on existing characters. There are several options: either the model "knows" them already or you could download a lora, specifically for a character you want to use. Another possibility would be using img2img though that might be problematic for non-fictional characters (consent wise). Here are a few examples that I created for you.
|
cDupin
Member
Posts: 476
|
So, I downloaded a new checkpoint yesterday bc I love to try out different styles. Here are a few examples I created in the last 10 minutes - just to show you that basically anything is possible. The only restraints are you creativity, patience and time:
|
cDupin
Member
Posts: 476
|
|
cDupin
Member
Posts: 476
|
And as I spotted a few annoying mistakes already: you can do several iterations of the same prompt and then chose the picture you like the most (that's how I do it for the Scarlett Blaze comic book series) or "fix them in post" (which I started doing recently as it saves a lot of time)
|
cDupin
Member
Posts: 476
|
Anyway, that's class for today, thank you! (and no day dreaming! you're embarrassing our teacher :tease 
|
vahtcpl
Member
Posts: 2110
|
nightbirds_2
Same question here
|
fred01
Member
Posts: 3404
|
cDupin: c'est le cours pour aujourd'hui, merci ! vous embarrassez notre professeur Top Merci pour tout et j'embrasse notre petite professeurs de partout, elle est plus belle que reel!! A tres vite pour la prochaine lecon fred
|
cDupin
Member
Posts: 476
|
vahtcpl: Same question here Hi, which question exactly? They have asked several questions and I have tried to answer them all. If I have overlooked something or if you've tried to set it up yourself but have encountered a problem, please elaborate. I am happy to help but I need more specifics, otherwise I'm clueless 
|
cDupin
Member
Posts: 476
|
fred01: Top Merci pour tout et j'embrasse notre petite professeurs de partout, elle est plus belle que reel!! A tres vite pour la prochaine lecon Elle sera ravie. Merci beaucoup.
|
sever
Member
Posts: 372
|
good ideas, i like generate my fantasy with ai, is quite simply but i find , until now, a loy of my fantasy in real wordl pivtures!!! but... will see .....
|
fred01
Member
Posts: 3404
|
cDupin: quelle est votre question exactement ? peux tu d apres cette photo, par exemple, creer un personnage voluptueux comme les tiens? simple essai fred
|
cDupin
Member
Posts: 476
|
sever: a loy of my fantasy in real wordl pivtures That is quite understandable. Personally, I always liked comics, particularly NSFW comics, so it's great to be able to do them myself (as I can't draw at all). But you are right, in this vast sea of porn out there, pretty much anything can be found. AI imagery is the perfect intersection for my reward system though - I get to learn new and exciting technical stuff while also getting horny 
|
cDupin
Member
Posts: 476
|
fred01: peux tu d apres cette photo, par exemple, creer un personnage voluptueux comme les tiens? simple essa Unfortunately, no. The only thing I might be able to attempt is "ai-ify" that picture. But a) it would be just this one picture and b) the more you "ai" it, the less it looks like the actual person and the more it becomes a generic looking ai character. If you want to change other aspects (eg. body shape or clothing) you have to "ai" it though. I hope it makes sense what I am trying to explain. The only thing you could do is create your own LoRA by using loads of example picutres and train the AI to make characters have her face. Alas, I have no idea how that is done, as I haven't looked into that yet. Sorry.
|
gregmartin
Member
Posts: 10729
|
|
gregmartin
Member
Posts: 10729
|
|
gregmartin
Member
Posts: 10729
|
|