I Love A Rainy Night (AI) [updated]

[La_Pucelle]

I thought change the theme.

I love! love! love! the color palette in this one. That red that's bronze that's brown - wow that's nice!

[Frank21]

Thanks LP glad you liked it. It's based on those 1960s Chinese revolutionary posters. I dislike doing portraits and don't really like the "wonderful" stuff.. surreal forest glades, dystopia, alien planets, futuristic cities etc.

[Paul (Fuzzy3000)]

@Frank22

@Legolas18

At present I am running things on a laptop with a Quadro RTX5000 - which has 16Gb.

To run things as a local install would I still need to install Python and GIT?

Not done a lot with AI yet [see previous reasons 🙂 ] but certainly looking at exploring more - especially if can be done for free.

So much out there to wade through to find the best methods.
Is there a decent Idiot's Guide that can be easily followed in order to get to know what to install etc.?

[Legolas18]

@Frank22 I love the new image! I don't think the last one was sexist, but it was sexy.

@Fuzzy3000

Just follow this guide from 2:22 🙂

[Paul (Fuzzy3000)]

@Legolas18

Thanks for that.
Pity there is not a written guide to go with it - sound and accent not the clearest at times [usually when talking about something unfamiliar - typical 🙂 ]
Oh well .... lots of pausing and rewinding.

[Paul (Fuzzy3000)]

@Frank22 Glad you liked that attempt.
I think your 1st images were within the realms of Tastefully Erotic - without being over the top.

@Legolas18 I have looked at both of the people you suggest - very comprehensive list of tutorials, especially Sebastion.
time to bite-the-bullet and get a local installation sorted.

You both produce some excellent work - maybe I will catch up soon, as well as having an extra option for the competitions... 🙂

[Frank21]

Last April I was pretty much in your position knowing next to nothing about AI and SD. It has taken about 6 months before I learned most of the ropes, though I could prompt a nice image close to what I wanted after a few days.
While the realism of the characters, lighting and clothing can look so much better than DS, you are left with multiple other problems in SD to solve. Not the least is coming up with an idea, concept or theme to elucidate in the first place, the same issue as DS or any form of art. I think I need another art comp for inspiration. Any takers, @Lapucelle?

[Paul (Fuzzy3000)]

@Frank22

Yes, I agree completely - creative juices [ooh-err] are definitely key to good results.
I suppose the main benefit of DS in that respect is that you can see more of the ideas by virtue of having the models to hand. You can also play around with poses, expressions, clothing etc. while keeping the exact same layout throughout the process.

Although there are some nice 'extras' available for SD that are promising for things like posing and Img2img.
Maybe one day we will have the same flexibility to add to or change even something we worked on previously.

[La_Pucelle]

think I need another art comp for inspiration.

@frank22 How about THIS...

NEAR or faaaaaar with DoF!

[diviniren]

This one definitely looks better than the original outputs. I think there's just something weird with the facial expressions and poses that doesn't sit quite right with me yet on some of these ai generated pieces. I think the issue has more to do with the inability to control the camera than the subjects themselves. There are certain poses that work real well with ai, namely mugshots and commercial/fashion type of poses. I wouldn't be surprised if this were the case because most reference pics are likely of modelling pics and advertisement/commercial content perhaps?

Really enjoy the evolution of your ai work @frank22.

[Frank21]

Absolutely correct and it's most evident that the images are from mug shots when you try to turn the faces even slightly side-on. When you're making a LoRa you're supposed to include a variety images including profiles but most don't bother to do this so faces will distort if you move them from the "looking at viewer" pose. I suspect that, particularly with celebrities, they just use use face-on Google images of them and nothing else. Imagine a world where where everyone's head is turned and facing you with a "please-love-me" smile.... shudder.... that's my vision hell.

[Paul (Fuzzy3000)]

OK - a little more experimenting, things starting to make a little more sense - just lots of it to try.
I now have a local install of SDXL, and played with a few settings.
Since I already have Harley Quin I thought I might as well experiment with Poison Ivy.
As usual, Rt Click to open in new tab for best viewing:

[Felicia]

Great images all - AI or not.

[Author]

Posts