|
|
# ? Aug 24, 2023 16:52 |
|
|
# ? May 29, 2024 23:52 |
|
WhiteHowler posted:I'm loving all the early AI stuff people are posting. The evolution is stunning. That is one big-rear end car.
|
# ? Aug 24, 2023 16:54 |
|
Food gone wrong, I was trying to get a moist beef sandwich and a can of soda I dunno where the hell the sushi rice came from. too dry! pixaal fucked around with this message at 17:27 on Aug 24, 2023 |
# ? Aug 24, 2023 17:13 |
|
Mercury_Storm posted:Got some high quality bathroom art now I guess you outpainted the original to get that? unzin posted:2 years ago I was playing with CLIP+VQGAN... can't find one working anymore, but I liked the abstractness of it. Coherent and Artistic modes are still available in Nightcafe. The former is just CLIP, the latter is VQGAN+CLIP.
|
# ? Aug 24, 2023 17:15 |
|
|
# ? Aug 24, 2023 17:47 |
|
That's a spicy take + arguably actually art what was the prompt for that
|
# ? Aug 24, 2023 18:44 |
|
a pilot seated at the cockpit of a 747, view facing forward of cockpit showing clear view, huge front window takes up whole front through forward window over shoulder of pilots view of two skyscrapers close up through front of window (region variation) view of two skyscrapers close up through front of window (region variation) square tall skyscrapers can't see top (region variation) the bay (region variation) over the shoulder view of George w bush piloting a 747 (region variation)
|
# ? Aug 24, 2023 18:56 |
|
Was that SD? Has anyone gotten the MIDJOURNEY region stuff to work? I can’t find it at all. Fake E: I don’t know why autocorrect put that on all caps but it’s better than constantly changing it to ‘midwife’ so I’ll take it.
|
# ? Aug 24, 2023 23:03 |
|
That was all using Midjourney's region variation. It shows up on upscaled images that haven't been panned or zoomed for me
|
# ? Aug 24, 2023 23:07 |
|
https://ideogram.ai/ Google (I think?) has opened this up, free to use AI image generation that's decent at text but is a bit uh, rough? A lot of the images remind me of Dall-E in the sandy, clay-like look some images would have. It's free and fast at the moment though so why not give it a shot?
|
# ? Aug 24, 2023 23:10 |
|
AARD VARKMAN posted:That was all using Midjourney's region variation. It shows up on upscaled images that haven't been panned or zoomed for me Hmm. I’m going to have to look at my settings. I don’t get the option to upscale anymore. I thought they removed it.
|
# ? Aug 24, 2023 23:17 |
|
Doctor Zero posted:Hmm. I’m going to have to look at my settings. I don’t get the option to upscale anymore. I thought they removed it. Maybe try doing a /prefer suffix Shouldn't require any settings
|
# ? Aug 24, 2023 23:20 |
|
KakerMix posted:https://ideogram.ai/ Keeps looping me back to the login
|
# ? Aug 24, 2023 23:30 |
|
Tunicate posted:Keeps looping me back to the login Are you not in the USA? I have a friend in Canada that was saying that it was dumping him back to the login as well
|
# ? Aug 24, 2023 23:33 |
Did the same to me, but the signup button seemed to work. Seems neat. Sedgr fucked around with this message at 23:39 on Aug 24, 2023 |
|
# ? Aug 24, 2023 23:34 |
|
Speaking of Dalle2, has it improved at all? I haven't even thought about it in a year and it seems like nobody here really uses it.
|
# ? Aug 25, 2023 02:25 |
|
I guess the business model is "it's too easy to scrape the entire Internet for the whole of human knowledge and feed it into a rented server for two weeks" followed by "I guess we'll just offer consulting services for our proprietary model " I guess the next step is to build an altruistic self learning AI that can't be trained on Nazi propaganda
|
# ? Aug 25, 2023 02:40 |
|
Really enjoy all the new MJ features. So much more control. Starting small and zooming out is great.
|
# ? Aug 25, 2023 02:42 |
|
feedmyleg posted:
Inpainting is what makes all of this for me. I don't know how MidJourney does it, but with Stable Diffusion you can inpaint whatever section at full-res which helps a ton with details and helps avoid the ~AI look~.
|
# ? Aug 25, 2023 02:50 |
|
Fuschia tude posted:I guess you outpainted the original to get that? Yeah I used IMG2IMG with a mask over the original a couple of times, and did some manual editing with GIMP.
|
# ? Aug 25, 2023 03:15 |
|
|
# ? Aug 25, 2023 05:44 |
|
sick!
|
# ? Aug 25, 2023 06:05 |
|
Haven't done a big post in a while, here's some things I've run off lately: Backgrounds: Phone wallpapers: A troubling puppet show: Bunch of deer: don't talk to me or my son ever again Some winged / horned skulls: Strange / slightly menacing weirdo figures: Spontaneous combustion: salty bois: The Pizza Wizard:
|
# ? Aug 25, 2023 16:32 |
|
Really nice. Prompt magic?
|
# ? Aug 25, 2023 17:30 |
|
RIP Syndrome posted:Really nice. Prompt magic? All done in Midjourney. A chain of describes / remixes, but it started with a /describe of Frida Kahlo's The Wounded Deer. That landed me with: a painting of a deer in the forest with arrows, in the style of frida kahlo, arthur tress, exaggerated caricatures, vincent desiderio, early medieval art, emphasis on facial expression, rasquache --ar 50:39 --upbeta --s 750 With a strong Vary applied and a weird filter of --w 2000 added, we land at a painting of a deer in the forest with arrows, in the style of frida kahlo, arthur tress, exaggerated caricatures, vincent desiderio, early medieval art, emphasis on facial expression, rasquache --ar 50:39 --upbeta --s 750 --w 2000 Some other honorable mentions and further variations on the same prompt with varying levels of weird: Also, here is a picture of a Pittsburgh bus sailing the open seas Edit: Inpainting in MJ combined with the weird filter is WILD. Did a describe on a Lichtenstein, cranked up the weird, then inpainted features in different styles to land on these: a painting with a woman head on the canvas, in the style of pop art: bold graphics, symmetrical asymmetry, jewish life scenes, associated press photo, mirror, celebrity-portraits, contrasting geometries --ar 56:67 --upbeta --s 750 --w 2000 Alternate path led to Escher-brains: Soulhunter fucked around with this message at 18:15 on Aug 25, 2023 |
# ? Aug 25, 2023 17:50 |
|
Alright, so I've been loving around on MidJourney and some online services like getimg.ai long enough. I want to be able to create my own dreambooth models. Does anyone have an easy install guide for M1 Macs?
|
# ? Aug 25, 2023 18:08 |
|
Thanks for the detailed prompts, I liked the deer one because of the nice balance between abstraction and realism, there are edges that hint at acrylic paint (I think, I know nothing about art), and it's not too Midjourney-y. Something else that just occurred to me is that the mouth/snout area somewhat evokes one of those stylized Guy Fawkes masks.
|
# ? Aug 25, 2023 18:10 |
|
|
# ? Aug 25, 2023 23:27 |
|
|
# ? Aug 26, 2023 00:43 |
|
The style I came in *intending* to draw: But I left in some realism camera style stuff in the prompt: I'm loving SDXL
|
# ? Aug 26, 2023 00:59 |
|
Try the crystalClearXL model https://civitai.com/models/122822/crystal-clear-xl and I mean with your same prompt /work flow and compare them pixaal fucked around with this message at 02:01 on Aug 26, 2023 |
# ? Aug 26, 2023 01:57 |
|
|
# ? Aug 26, 2023 03:11 |
|
Soulhunter posted:A chain of describes / remixes, but it started with a /describe of Frida Kahlo's The Wounded Deer. Is this describe-then-generate feature a one-click operation anywhere? Very interesting things could be done with this to explore the feature space or just do weird stuff.
|
# ? Aug 26, 2023 03:38 |
|
|
# ? Aug 26, 2023 04:49 |
A couple of fish, a disconcerting flower, some nice crystals and a very purple baphomet.
|
|
# ? Aug 26, 2023 14:44 |
|
Here's a prompt challenge for you (without img2img trickery). A djinn emerging from a magic lamp, only the torso is visible, the lower body is obscured by a swirling cloud of smoke. (Not the actual prompt, just a briefing, basically). I can't get it to work, neither by trying to evoke the mere concept (arabian nights tale, aladdin, magic lamp, djinn) nor by trying to verbalize it without any contextualization. Even negative prompting of "legs, lower body" does not help, it's always a full body. Super weird. Tried with multiple checkpoint model types. Some pretty results, but not what I'm looking for.
|
# ? Aug 26, 2023 16:16 |
|
Had to do it piece by piece with outpainting in MidJourney, but not hard after starting with the face. Had to rerun a few with negative prompts for "fabric" and such when doing the smoke. But yeah, even plugging the final result back in with an image prompt wasn't successful, it doesn't seem to have a good sense of what a djinn is. Between this and the mermaids, it seems to have problems with hybrid creatures.
|
# ? Aug 26, 2023 17:20 |
|
Would genie work better than djinn?
|
# ? Aug 26, 2023 17:22 |
|
It didn't do great with the "lamp" part, but I was able to get a djinn/genie with an incorporeal lower body.
|
# ? Aug 26, 2023 17:55 |
|
|
# ? May 29, 2024 23:52 |
|
I tried all the spellings separately as well as together, nada
|
# ? Aug 26, 2023 17:56 |