Dall-E, Midjourney, Stable Diffusion (etc) - who's playing, and with which?
- Started
- Last post
- 894 Responses
- zaq3
- https://stable-diffu…Nairn
- (this is gonna be a pivot of some thriller movie.. whiz detective lies down in the outline of the murdered.. voilà. and AI is stopped in the nick of time)Squiddy
- Does anyone actually know how to follow that tutorial? I feel so old trying to learn all this stuff. Fuckit.HAYZ1LLLA
- You should listen to some Radiohead and let it guide you, @HAYZ1LLLANairn
- I bloody love Radiohead me.HAYZ1LLLA
- uan0
Uploaded screenshot top left of a line drawing from illustrator, and pressed some promeai.com preset buttons.
tool linked by phanlo.
prompting in midjourney with same screenshot
- PhanLo4
Here was an attempt using that https://www.promeai.com site tonight. The guy I work with often paints giant skeletons and we're trying to work out a wall where it's a big forest jungle and lots of graffiti is mixed in it but there will be a slightly off-centre piece skeleton/person to break up the compososition.
So I rendered the skeleton as lines using freestyle in Blender then just added a simple vector graphic for the leaves I found online
-
I only found out Blender did the outlines thing the other day, honestly wish I learned it years ago.
-
Right anyway, here's what it started spitting out
I haven't registered with the site and just download them as I go by right-clicking and clicking open in a new window, for some reason it downloads it as a higher resolution than downloading it the other way. I generated 258 images
-
-
That's a short selection, I probably got something close to what I wanted in the first 20 images but I just hammered through a bunch of variations to see how off it would go.
Then I took the images into Photoshop and extended them out with the generative fill, I think doing incrementally gets better images but the last few I just did in one go.
-
Weird that this didn't exist, at least for me a week ago, but it's a workflow I would use 100% to paint a wall. The skeleton I could render in Blender properly and comp it into the design with exact lighting and the background I'd redraw anyway, but some decisions are helped along.
-
I share a lot of this stuff with other painters, but only a couple of them seem to get it. I mean a lot of us are close to 50, some guys paint more or less the same thing that they perfected as a teenager, just with a different fade, so it's too late for them., their best days were 30 years ago.
-
A guy was in the studio today talking about controlling live feeds from projectors with AI so you could monitor the output remotely which is pretty handy, so you could live update content offsite.
He's an interactive programmer who builds a lot of custom triggers for activations, triggering content or lighting in venues, it was interesting to hear what he thought of som of it.- I was downloading as I was going because I realised if I didn't the images started to disappear because I wasn't logged in.
just in case you get images you likePhanLo - when i use sketch to image it just spit out different sketches. what am I doing wrongwhere_am_i
- Did you go to this one? https://www.promeai.…PhanLo
- yeah, its doing the opposite for me. its rendered sketched version of an image or sketch, not photo from sketchwhere_am_i
- ah weirdPhanLo
- Thats really interesting thanks for sharing.
looking at it yesterday I thought it could only steer towards fairly set generic styles - liek arch-viz orwoowahesque - manga character but it looks much more versatile and responds to the lines nicelywoowahesque
- so if you upload sketch it makes you a photo type image?where_am_i
- oh i seeee - change scenario up topwhere_am_i
- I was downloading as I was going because I realised if I didn't the images started to disappear because I wasn't logged in.
- grafician-1
"Comparing Adobe Firefly, Dalle-2, OpenJourney, Stable Diffusion, and Midjourney"
- SD is basically the best rn
I guess most people can't distinguish generated images from real photos at this pointgrafician - Really? I don't like the versions here. Impressed how far FF has come, though.formed
- yep. every example here has AI generated all over it.uan
- The last one is SD, looks very "normal realistic 3D rendering"
The rest indeed look very "AI generated"grafician - just look at the shadows. they aren't normal. nor is the architecture.uan
- Don't need to be, as an interior designer you can get a lot of decent concepts even sografician
- SD is basically the best rn
- grafician0
- This leads me to believe those "zoom and enhance" scenes on TV are gonna be real, real soon.robthelad
- +1 For animations I think it will be a really decent feature, all of the ones I've tried have been good.PhanLo
- fake out-paintingutopian
- it just keeps getting darker to the left as the driver heads towards oblivionsarahfailin
- Nairn2
This is quite a neat sequence..
- PhanLo0
text to video nightmares
https://huggingface.co/spaces/da…
-
- PhanLo1
- not sure I like the look of the mustard https://i.imgur.com/…PhanLo
- What's your prompt in Photoshop? I get random sausages on a dark table.stewart
- I think I had 'glistening photorealistic sausages on a plate' possibly. Took a few go'sPhanLo
- It's not following the QBN shapes at all here. Interesting!stewart
- Was a bit hit or miss for me too, just tried https://imgur.com/a/…PhanLo
- Haha cool, that's not what I get though.stewart
- _me_2