Artificial Intelligence
- Started
- Last post
- 1,637 Responses
- palimpsest0
Didn't know this would turn out to be a flex but my shit's so wack no AI is claiming it.
XD
- shapesalad0
- I just got "months until photorealism" for what's next.
OKgrafician - yep, i caught that. Expect video/motion/animati... by mid 2023 i think, and then job centre for new career...shapesalad
- video motion animation on what hardware?
lol you can barely run the txt2img models on latest nvidia cardsgrafician - ^was talking about stable diffusion that can run locallygrafician
- I just got "months until photorealism" for what's next.
- grafician-2
Btw, Stable Diffusion got over 1M users already in what, like a week since open sourcing it?
- https://user-images.…kingsteven
- Wait until they allow nudes!shapesalad
- kingsteven7
i've had stable diffusion running on my mac in the background for a couple days - generating an image every 45 seconds at 640 x 512 px. i've been occasionally changing the prompt to see what works. copyright arguments aside it's nice to land in a pocket and watch it do it's thing
here's 100 sequential images from a single prompt "lomo ghost in crystal light refraction" from this evening upscaled to 4k
- Yeah how is that not awesomenb
- Better than Netflixnb
- incredible. far more interesting than the usual AI artinteliboy
- How much longer does it take to render if you double the resolution?drgs
- it glitches out on me over 640px, i think the sample image size is 512 x 512 so not sure how i'm even doing thatkingsteven
- see the video i posted in the sidenotes below. A.I. upscales are decent, in combination with out-painting and image-to-image you can get hi reskingsteven
- i noticed over 45 seconds my images were mostly garbled or black (1 in 5 would get to 2 minutes and fail) so i set a time outkingsteven
- i've added another load of 100 images with the same prompt to the gallery, astonishing variation when you keep it vague https://imgur.com/a/…kingsteven
- ...and find something it likes drawingkingsteven
- 512 is standard, you can also choose 3:2 or 2:3 (portrait) but it will cut and paste very crudely elements of the image
ex: https://i.imgur.com/…grafician - Very cool. I'm seeing a lot of '90s shoegaze album covers in amongst the nightmares.MrT
- It's cool and I hate it.garbage
- Dunno which fork you're using but if the AUTOMATIC1111 version runs on mac, there's a highres fix option that allows for coherence at high resolutions now.kalkal
- Two step process, it creates a low res latent noise then uses that to inform the higher res version.kalkal
- Beeswax0
Midjourney added tiling feature.
Now you can create seamless patterns using --tile
Not perfect but satisfactory
- utopian2
Researchers Say It'll Be Impossible to Control a Super-Intelligent AI.
- drgs2
People are submitting AI-generated images to Shutterstock/Getty images. So what happens when next generation image AI scrapes these AI-generated images and learns from them?
- we finally get video feedback art in color!uan
- "Getty Images bans AI-generated content over fears of legal challenges"grafician
- https://www.theverge…grafician
- "Getty Images CEO Craig Peters told The Verge that the ban was prompted by concerns about the legality of AI-generated content and a desire to protect the site’grafician
- ^ told you there were copyrights issues with all AI generated content :)grafician
- gaitty imagesshapesalad
- That's what DALL-E, Midjourney, etc are for. They're GANs. The current human input is forming a discriminitive library bank..garbage
- ..eventually the "writers" will not be necessary. The "A" in GAN stands for Adversarial. The whole point of the system is total automation.garbage
- Said it before, said it again: If you're in your mid-30s and spent your entire life learning Adobe, bad fucking news for us.garbage
- Basic things like Adobe Express? Guarantee you that data and feedback is going to Sensei.garbage
- There will be a massive "FUCK ADOBE" thread surge a few years down the road whenever they unveil whatever monster they are working on.garbage
- I asked a getty rep in a meeting about this very thing five years ago and the bitch laughed, brushed me off, and gave me a promo thumb drivescarabin
- garbage, I get the idea. but what do they need the human input for? they can feed the machine with art history / reviews already.uan
- I'm guessing most of these libraries already have a baked-in Louvre, but why not have a bigger library for what is currently trending?garbage
- Neither side uses the brakes in a zero-sum game.garbage
- grafician-2
Nvidia Reveals RTX 6000 With 48GB GDDR6 ECC Memory
"Nvidia says that its RTX 6000 graphics card with 48GB of GDDR6 ECC memory onboard will offer up to 2X – 4X the performance of the previous-generation RTX A6000 due to the massively increased number of CUDA and RT cores as well as the brand-new Ada Lovelace architecture."
- shapesalad2
All these AI image generators.. I presume these images are not layer.
It's going to be a world of pain when your client gives you an AI image to work with...
- *layeredshapesalad
- no layers
it's basically "logo in a Word .doc"grafician - they'll be an AI script you can run to separate into layers. AI will fix all the problems it creates.dopepope
- Uhhh dude this software has been out for like two months? Give it a minute more and it’ll be creating infinite layer comps for yanb
- wait for itApeRobot
- grafician1
database of prompts
- nb1
- grafician-2
Anybody running SD locally on 32GB M1 Max?
How fast? 64GB better?
- i read in early M1 tests the 64GB was 10 times faster than the 16GB, no idea about 32. about 45 secs on my 64 for good quality but i've no idea what im doingkingsteven
- <1 min seems fair on current local hardware, thanks I'll give it a trygrafician
- yeah, and gets comparatively better with larger models (that would otherwise require 24GB RAM on dedicated graphics hardware)kingsteven
- Ramanisky20
- It seems like every day is a 'this changes everything' momentIanbolton
- NBQ000
I can't get StableDiffusion (0.2.0 or 0.3.0 versions) to work. Every time I double click on the CCX file it says it requires compatible app.
I'm on a Intel iMac Pro with latest Adobe CC.
Any ideas how to get it to work for Photoshop?
- Story of my life.
I hate fiddling with setups.palimpsest - google how to install photoshop plugins manually :-skingsteven
- FYI i'm assuming the 'stable diffusion plugin' requires stable diffusion to already be installed + running on your mac... it may be very slow on an intelkingsteven
- and you should really be more careful downloading and double clicking things you don't understand :Dkingsteven
- I thought the official download from Chris Cantrell site is just the CCX file and it's all you need to make it run, no?NBQ00
- Delete System32palimpsest
- ahh, it's just an API plugin - i'd never heard of it. it requires photoshop 23.5.1 - installed fine for me once i upgraded.kingsteven
- i guess this would be akin to an advanced online content aware fill. this will be obsolete very soon. multiple editors in the works with A.I. toolskingsteven
- Story of my life.
- sted0
- is this a thing it does inadvertently? i've noticed the images it makes have small file sizes.kingsteven
- I would guess the position of the pixels is mathematically better aligned for compression. because it 'places' them instead of interprets them from input.uan
- @kingsteven inadvertently i don't think so, as i understand its a byproduct of composition optimizationsted
- ah, that's kind of what i meant. i was wondering if this was a specific technique for compression using SD or just a byproduct of reprocessingkingsteven
- NBQ00-2
How long until you can type prompts in Blender and it would create entire 3D models with textures and everything and then you could even have it animated?
On top of that you could define a desired style and it would create and setup the entire scene appropriately?
- NBQ000
I know AI (prediction models) are being used in some ways in the stock market, high-speed scalping and so on but it's not able to give accurate results or the hedge-fund owning the tool would be Trillionaires in no time.
But what happens to the finance markets the day an AI can accurately predict 60%-80%+?