The poor things must have gotten deformed in the miniaturization process.

Or I could just use Photoshop along with decent source images and create my final image the oldfashioned way, from scratch. This hideous AI monstrosity is so bad...the second woman from the left has no arms!Prof Sai wrote: ↑Tue Jan 16, 2024 5:29 amWhat I would do here is zoom in on the individual figures, and then use image-to-image to re-render them. If a face is in closeup, the AI generally does better. Then you have to go into an art program, upscale the original, and then fit the new images into it. They usually won't match exactly, so this takes some work with layers, and the erase and warp tools.
When you worked for hours on a gen going through prompting, then inpainting/outpainting, refining and trying to fix everything, then post-process the whole thing in Photoshop/GIMP and someone comments "hurr durr, mAGiC cLIck mAdE uR piCs".
Interesting views.Raso719 wrote: ↑Tue Jan 16, 2024 3:23 pm
Both are tools and as someone who works in an architecture firm and manages tech and software there both photoshop and AI are fantastic tools for producing multiple permutations of visualizations but both have a learn curve and both sort of need each other to cover each other's asses.
Yea, I know. Or they just announce that anything AI is "soulless" on principle. I've had some dismal artists do that.foreverlurk wrote: ↑Tue Jan 16, 2024 3:32 pmWhen you worked for hours on a gen going through prompting, then inpainting/outpainting, refining and trying to fix everything, then post-process the whole thing in Photoshop/GIMP and someone comments "hurr durr, mAGiC cLIck mAdE uR piCs".
So does that mean AIs are ginger?
Little Sally wrote: ↑Wed Jan 17, 2024 9:30 amPs. Serious question. Having a liking for creating the occasional bizarre fantasy pic, what is AI like at creating tentacles? Anyone tried it yet? (talking about the trope of women in the clutches of such things etc)
Hey that looks very promising, thank you! (the right hand looks a bit odd but could easily be corrected). Nice facial expression and body shape, and It has a very dramatic look to it overall.who renders 3d wrote: ↑Wed Jan 17, 2024 4:35 pm(image)Little Sally wrote: ↑Wed Jan 17, 2024 9:30 amPs. Serious question. Having a liking for creating the occasional bizarre fantasy pic, what is AI like at creating tentacles? Anyone tried it yet? (talking about the trope of women in the clutches of such things etc)
It's nice you've had a go though. Is it hard work to get the text prompts right? Or is it this censor thing I keep hearing about.
who renders 3d wrote: ↑Wed Jan 17, 2024 4:35 pmLittle Sally wrote: ↑Wed Jan 17, 2024 9:30 amPs. Serious question. Having a liking for creating the occasional bizarre fantasy pic, what is AI like at creating tentacles? Anyone tried it yet? (talking about the trope of women in the clutches of such things etc)
![]()
I have a local generator so I don't have to worry about censorship. But there are all sorts of models and LoRAs and control settings that you need to understand, even before you start doing prompts. And I'll admit I didn't really try that hard, since the topic doesn't interest me.Little Sally wrote: ↑Wed Jan 17, 2024 6:22 pm
It's nice you've had a go though. Is it hard work to get the text prompts right? Or is it this censor thing I keep hearing about.
Again, a nice looking image, thanks! (she looks like she's enjoying the company lol!)
Understood, that's fair enough.Prof Sai wrote: ↑Wed Jan 17, 2024 7:03 pmI have a local generator so I don't have to worry about censorship. But there are all sorts of models and LoRAs and control settings that you need to understand, even before you start doing prompts. And I'll admit I didn't really try that hard, since the topic doesn't interest me.Little Sally wrote: ↑Wed Jan 17, 2024 6:22 pm
It's nice you've had a go though. Is it hard work to get the text prompts right? Or is it this censor thing I keep hearing about.
Yes I get that Who, and 3d is what you know best eh.(good image BTW!)who renders 3d wrote: ↑Wed Jan 17, 2024 7:20 pm(image
I don*t know, its trial and error. I would rather build a 3d scene instead of rolling dices to get 1 good picture out of 50.
This is the biggest drawbacks with generating from scratch right now. It's why in my office we use it for generating permutations of existing designs instead of generating new ones.who renders 3d wrote: ↑Wed Jan 17, 2024 7:20 pmI don*t know, its trial and error. I would rather build a 3d scene instead of rolling dices to get 1 good picture out of 50.
I did that with a european city .. it turned into something build from lego with lego people on the street.Insanity.Raso719 wrote: ↑Wed Jan 17, 2024 7:46 pmThis is the biggest drawbacks with generating from scratch right now. It's why in my office we use it for generating permutations of existing designs instead of generating new ones.who renders 3d wrote: ↑Wed Jan 17, 2024 7:20 pmI don*t know, its trial and error. I would rather build a 3d scene instead of rolling dices to get 1 good picture out of 50.
Heh! Just been having a play with that freebie Microsoft Bing image creator. Not had much luck with the tentacles yet, but it threw this one up that amused me. I'd pay to see this act lol! Love that leery expresion on his face lol!Prof Sai wrote: ↑Wed Jan 17, 2024 7:03 pmI have a local generator so I don't have to worry about censorship. But there are all sorts of models and LoRAs and control settings that you need to understand, even before you start doing prompts. And I'll admit I didn't really try that hard, since the topic doesn't interest me.Little Sally wrote: ↑Wed Jan 17, 2024 6:22 pm
It's nice you've had a go though. Is it hard work to get the text prompts right? Or is it this censor thing I keep hearing about.
I guess you will never see a tentacle around anyone with Bing, but you could add something like "photo","realistic","no painting" to your prompt. That is the key, you have to tell the AI exactly how it should look like.If you want a comic book style .. "no photo","cartoon" .. AI can create nearly anything .. just no shrunken women and tentacles out of the boxLittle Sally wrote: ↑Wed Jan 17, 2024 10:07 pm
Heh! Just been having a play with that freebie Microsoft Bing image creator. Not had much luck with the tentacles yet, but it threw this one up that amused me. I'd pay to see this act lol! Love that leery expresion on his face lol!
Then there was this one I thought was rather nice.
Ha ok, good tips! I was just having a play and it's quite good fun to see what you get. I'll probably have to buy my own giant octopus and train it to do tricks!who renders 3d wrote: ↑Wed Jan 17, 2024 10:37 pmI guess you will never see a tentacle around anyone with Bing, but you could add something like "photo","realistic","no painting" to your prompt. That is the key, you have to tell the AI exactly how it should look like.If you want a comic book style .. "no photo","cartoon" .. AI can create nearly anything .. just no shrunken women and tentacles out of the boxLittle Sally wrote: ↑Wed Jan 17, 2024 10:07 pm
Heh! Just been having a play with that freebie Microsoft Bing image creator. Not had much luck with the tentacles yet, but it threw this one up that amused me. I'd pay to see this act lol! Love that leery expresion on his face lol!
Then there was this one I thought was rather nice.