The first 1,000 people to use this link will get a 1 month free trial of Skillshare: skl.sh/bycloud07212 Have fun learning well organized courses for 1 month free!
ah yes and V - Taming C - Learning Q - Transformers for L - Transferable Visual Models G - High-Resolution I - From Natural Language A - Image P - Supervision N - Synthesis
I dunno about VQGAN, but CLIP is actually "*C*ontrastive *L*anguage-*I*mage *P*re-training". EDIT: Yes, the asterisks were meant to embolden the text... but whatever. You get the point.
Gosh, I just love the slow and steady improvement AI art has gone through in the past few years. The sheer level of experimentation and trading and melding of techniques makes this feel like a real underground art scene. And it's getting better all the time!
Had so much fun with this tonight! Underwater city | plane crash | steampunk made a beautiful scene. Thank you so much for these video's, as you'd be able to tell from my channel, I think highly of AI.
Hey Bycloud, thank you for sharing with us, how can I stop the VQgan Clip distorting my original photo so much right from the start ? I want the first image of the video to be clear, not an 20% distorted image.
If anyone wants the pictures displaying the results produced by specific keywords shown from 1:40 to 1:53, it's in this tweet ! 👉 twitter.com/E08477/status/1414442147877695488?s=19
2:24 "click this and click that, then click restart and start clicking again" ? wtf man are you trolling ? i mean did you really have to f*ckin speed run the most important part of the video ..
Bro, with VQGAN+CLIP you really don't need any more than 200 iterations. Any more and you're just wasting time and holding up the GPUs that others could be using, because your image isn't going to get any better beyond a certain point..
The first 1,000 people to use this link will get a 1 month free trial of Skillshare: skl.sh/bycloud07212
Have fun learning well organized courses for 1 month free!
ah yes and
V - Taming C - Learning
Q - Transformers for L - Transferable Visual Models
G - High-Resolution I - From Natural Language
A - Image P - Supervision
N - Synthesis
Bababooey
I dunno about VQGAN, but CLIP is actually "*C*ontrastive *L*anguage-*I*mage *P*re-training".
EDIT: Yes, the asterisks were meant to embolden the text... but whatever. You get the point.
@@KoolenDasheppi Well, the GAN part is probably Generative Adversarial Network, so now we just gotta figure out what VQ means.
@@desu38 Vector Quantized Generative Adversarial Network! I got it! Hell yeah.
@AI Masterye🎉
Gosh, I just love the slow and steady improvement AI art has gone through in the past few years. The sheer level of experimentation and trading and melding of techniques makes this feel like a real underground art scene. And it's getting better all the time!
Have you seen Dall-e? It’s vastly far ahead of this, just not open to the public. Check out the interactive demos on the site.
Holy moly, those are some INCREDIBLE results.
this is what fortunetellers see in their crystal ball when you ask them
Had so much fun with this tonight! Underwater city | plane crash | steampunk made a beautiful scene. Thank you so much for these video's, as you'd be able to tell from my channel, I think highly of AI.
i love these images more than i can say and i find it amazing and terrifying at the same time that a machine made them
Use this visual code on the library of Alexandria 👀👀👀👀👀👀👀👀👀👀
looking back this feels like the pre-historic times..
I can't believe it.. I WAS ON TVVV!!!
Hoho
Huhu
now its more easy to make games about man who slowly goes insane... this is for pictures in rooms are good
2:07
Sus
Dude! This is wild!!!!!!!
That sus image is cursed
Cursed af... nightmare fuel
Does it matter what notebook I'm using? I keep playing with this but it keeps giving me crap images lol nothing like these in this video. Any insight?
It's unreal! ))
I do not see a notdeepdream subreddit under deepdream. is there a link?
Hey Bycloud, thank you for sharing with us, how can I stop the VQgan Clip distorting my original photo so much right from the start ? I want the first image of the video to be clear, not an 20% distorted image.
If anyone wants the pictures displaying the results produced by specific keywords shown from 1:40 to 1:53, it's in this tweet ! 👉 twitter.com/E08477/status/1414442147877695488?s=19
Anyone can be an artist now.
Thank you!!!
Hmmm I don’t get it. Is it an open source model? Like smth built on tensorflow? Or is it a closed source AI generator that you can only call via API?
i love these trippy ai art
which model do you think its best
This is insane.
I tried running it but it says "overflow" i guess my computer can't run it.
nvm i tried a different one and it's working
That is surreal stuff. Imaging using your name as a promt and it generates your face lol
Lel
And it has a smile on the face too-
who is the user that came up with the test for keywords? at 1.34
Never have I clicked it so fast!
Anything new about anime :D
👍🏽
When you say “content” in the phrase “hearts content” then you need to put emphasis on the last syllable.
It’s con-TENT.
Not CON-tent.
1:40 pixiv!
NFT anyone ? Lol drowning in money
2:24 "click this and click that, then click restart and start clicking again" ? wtf man are you trolling ? i mean did you really have to f*ckin speed run the most important part of the video ..
So basically, the AI creates garbage?
Dude, I'm a real human and the T2S program they're using for these channel's vids is a pain to listen.
Bro, with VQGAN+CLIP you really don't need any more than 200 iterations. Any more and you're just wasting time and holding up the GPUs that others could be using, because your image isn't going to get any better beyond a certain point..