Tillerz
Tillerz
Vladmantic added a second pass to the regular process ... the "refiner" further down is for SDXL.
Tillerz
It also eats less memory now and is faster, too.
Tillerz
https://imgs.plurk.com/QD1/hzq/zmdCxzV2XUF88NfJpzgyPhAoXkt_lg.png Also I used Dreamshaper 7 model.
Max Graf
you are both on another level with that stuff. I had no idea the input/prompt was that involved.
Tillerz
If done right you can control a lot of the stuff you see.
Tillerz
It's getting bits and bobs left and right, not just cloning an original image and then adding a style ontop.
Tillerz
Max Graf : https://images.plurk.com/2BRcRMhfqK0GDakjcFLej2.jpg The prompt for this was:
Tillerz
It didn't get everything in, like the cyan ponds, but if I had chosen a panorama format, it probably had fit those in, too.
Tillerz
But you have the fae like woman, the blue hair, the collar, the lots of jewelry, the castle and the plants, also the "Alphonse Mucha" style stuff on the ceiling and the windows, and the god rays... :-D
Tillerz
...the frontal light and the very detailled skirt.
Tillerz
I want to see the artist who can say "this is a copy of my artwork" ... :-)
Tillerz
I wonder if someone will make an AI at some point that will analyze an image and find the closest "original" thing, or even multiple if an AI image is 'put together' from several originals.
Tillerz
Google lens is rather dumb right now, they should plug in an Ai there.
Tillerz
But still with making AI images, the "art" right now is to not get the "stupid stuff" but make it actually do what you want.
Tillerz
So nothing compared to what artists do, it's more of a "language" problem.
Tillerz
Max Graf : That image at the top though, it only got VERY BASIC instructions.
Tillerz
Says nothing about the content, only about the quality and the style.
Tillerz
That kind of prompt is not helpful if you want to generate unique characters that you can repeatedly use for different images, cause it will be a different person every time.
Tillerz
"(thirty years old french woman Maarika Sujatha with (long hair black updo hair):1.2)" this is enough for a unique character. https://imgs.plurk.com/QD1/ORj/fRLtLP6XMkdVdlaS6v6dPYjgcSj_lg.jpg https://imgs.plurk.com/QD1/MsQ/bjqL0Wuk3pOH3YyQWmCjRVzysrJ_lg.png https://imgs.plurk.com/QD1/FBk/71MRqp4zBtkhF7vDpFTTqK5erYf_lg.png
Tillerz
Parts of 3 images in a row, the sama character. :-)
Tillerz
The "origin" (that's what I call it) also defines a bit what shows up in the background, if you do not describe it properly.
Vic Lenoirre
Nice!
Tillerz
So if you have an "egyptian woman... " ... and no background description, you will most likely see some pyramids in the background. :-D
Tillerz
There: https://imgs.plurk.com/QD1/v9G/OUXwmb67IOlMC2yGyX83dYASnfO_lg.jpg Prompt was ONLY "egyptian woman Nofretete Hamadi with black bigtails in a white dress, outdoors"
Tillerz
Added a little bit and voila: https://imgs.plurk.com/QD1/fAB/tlQ3d0cqaeIMUntGrVGkzjZRQs9_lg.jpg All I added was ", in front of a norvegian ice sculpture, in winter, at midnight"
Tillerz
And if you like Alphonse Mucha, just add it (in the style of alphonse mucha): https://imgs.plurk.com/QD1/yuT/DcuHav5WPNAj7lzlc0PA6VPfaCF_lg.jpg
Tillerz
And if you want egyptian ice sculptures made by Picasso, go at it. :-D https://imgs.plurk.com/QD1/TxD/axEIWKsP6gbaTYcRLmEsOYEvkmj_lg.jpg
Tillerz
Also that only takes like 12 seconds to render for a 1024x0124 photo. So people can do shit really quick.
Tillerz
And you can upscale it too. 2048x2048 here: https://images.plurk.com/2o7kcaspA8mEVmZRreH9Pn.png
Tillerz
The ice sculpture loses a lot of details, but the woman is way more detailed.
✴Sparkling✴
these are amazing
Exacerangutan
my favorite part of the first Hamadi one is the surfing wine bottle
Tillerz
Tillerz
https://images.plurk.com/40Py2pIxPBtCNCN57KYoDw.jpg
Exacerangutan
AI: what about boobs
me: I for one welcome our new AI overlords
Tillerz
載入新的回覆