I've been thinking the same thing. I predict that in 10-15 years this uncanny AI will become its own niche art form with a dedicated community of hipsters. It will be called AI-core
@@aurin_komak I already love the look of abstract AI images because they look like my dreams so when AI is disgustingly realistic and takes my job as an artist i hope they remember the uncanney valley era we were in
You should really talk about DALL-E 2. It's still in beta, but holy crap you need to see the progress they've made. It's honestly mind boggling. I do believe there's a really long wait list, but it's worth checking out. A huge step over this DALL-E version. Open AI really seem to know what they're doing
it's actually amazing, I keep trying to stick funny characters into real life locations and it's terrifyingly good. it even got a good render of my home town or the village I live in now, and all look accurate for the type of housing around here. only problem for a lot of people is there are a lot of limitations on the new ones, the worst being 'star WARS'. Its a lot of fun for waluigi pictures though.
Open ai is a bad company. The only difference to dalle Mini is that opn ai has mor gpu and bigger data Sets that's all. The quality of the immaes Skalen in relation to training power and dataset size. Open ai does nothing special. Just look at Googles Text to immage ai it's xalled imagine. The quality is even better then dalle2. Because they have even bigger data Sets and even more gpu. At the moment ther is no real scientific progres in this area in ai. And the worst thing is that the big ones dosnt get release to the puplic. Dalle Mini is developed by a community and not by a bad company. And they release the source Code and the training Checkpoints. Because of this dalle Mini is better
I find it works incredibly well when you give it an art style, in particular "photorealistic", "courtroom sketch" is incredible, "funko pop" of x is also great
Also some conceptual situations like a doll- or crochet-version of objects or people. Muppets look good and movies work well as templates, as well as "trail cam footage".
This AI has no chill. I asked it once "show me how you see us humans" and in response I got 1 image with one half showing a monkey and the other half a really dumb looking depiction of a human, almost as if it was mocking how humans look ...and the rest was all skulls D: ...stopped asking for pictures after that
This ai is incredible, unfortunately it's gotten super popular lately, meaning it's unlikely to work as the traffic is too much for the servers. Hope they increase their servers or make their own website or something.
@@KingLich451 Still preferable to groomers. Detransition accounts are absolutely harrowing and the people responsible just up and washing their hands of the matter need a floor to ceiling bath.
You should try and apply for Dall-E 2, I bet you could get a version to test if you provide how you're a tech youtuber and often talk about similar stuff.
That's not even the most advanced model openai has. DALLE-2 is right now only available to researchers and let me tell you, it's something special from what I've seen on the internet. Altho the pics shown were probably curated.
The most interesting thing about DALL-E Mini is that the model is completely different to DALL-E 2 and slightly inferior even from DALL-E 1. Plus there are no restrictions on people or faces that DALL-E 2 has so arguably has more use cases for the general public. It's funny how this has become a lot of online people's first introduction to AI and in a way I'm glad it's this lesser model without restrictions. It gets people talking about the ethics of using real people in completely faked images while not having a high enough quality to be able to trick people. Sort of like a training ground for the general public on AI. Plus if everyone had access to DALL-E 2 I think people might have been underwhelmed by some of those restrictions anyway since DALL-E 2 is best at creating the sort of stuff that doesn't spread as quickly in "Meme Culture" circles.
You're completely correct, the more details you give it the more it knows what to make specifically, if you give it something general and abstract it makes a wide variety of things all mixed together and thus looks worse. Based on interpretation it can create a wide variety of novel contextual images but you should be specific about the context if you want a clear conclusion
The Smaug as a beanie baby itself was awesome! I put in Freddie Mercury drinking tea, and it seemed to grab information from live performance photos of Freddie, and slap a tea cup in his hand instead of a mic. I was not disappointed, lol.
It also helps to specify the style you want it to output. I came up with some awesome results for "farewell, expressionism" and "spider, da vinci" or "dog warrior, pencil". Improves the results by a lot!
the strangest thing is that after generating these frightening faces, I had such a feeling in my eyes that everything is crooked like in these photographs.
What I find interesting is just how dreamlike these images look. Its understandable, as its basically the same idea to dreams, you take a lot of inputs and reconfigure them to fit the currend demand. And as our sense of reality is turned off during dreaming, it gets very weird. It has also similaritys to how things look if you are on a mushroom trip. And it makes one wonder if its a similar mechanism, or if its completly diferent but arriving at the same result.
From an artists perspective, I would use this to help with just thinking of what to draw and using it as a sort of underpainting or thumbnail sketch to work off of.
Giving the prompt generator HIGHLY specific words like refractive, talking to, rolling down, also helps bring two objects to life within each other. My first two were "Wizard talking to Amish people" and "Hypersphere rolling down a mountain". I got some fascinating results, some were far less so than others such as "tornado wearing sunglasses" (none of them hit the mark, one of them was a cartoony alien head, but the actual tornados themselves were fairly interesting). Getting too specific however, would just end up with it focusing on something extremely easy, like "anomalous dark spots within a tesselated garden" (pardon spelling mistake) just go me this black weirdness. Here was my foray into it seeing a random twitter link in duckduckgo that I mysteriously clicked on that was totally unrelated to my search, it was fun. (chronological order): Amish | i.imgur.com/bhP5OOz.png Hypersphere | i.imgur.com/QDl9CGf.png Unseen truth | i.imgur.com/CXp1N6K.png (was curious to see if grammar would do anything, and then I saw the beaches and decided I wouldn't think further on it.) Tornado | i.imgur.com/9NB8NLy.png Room | i.imgur.com/R9fxwbE.png Dark | i.imgur.com/iMaIMFQ.png Coins | i.imgur.com/7wUfB2D.png Texture | i.imgur.com/GdeJhyn.png A-pylon-clipse | i.imgur.com/Zvi2Ayw.png Fruit | i.imgur.com/8T21Svc.png
I love these types of things so when I saw it in ur vid I knew it would be worth my time. For whatever my mind could only think of "Kratos eating a cheeseburger next to the leaning tower of pisa" and to my surprise it did it. I have no clue how they do it but I love it.
Also, in case you read this which I'm not expecting, try Latent diffusion. I believe you can Google "LAION 400M Google colab" and you should see it. Although you need to get more creative with the prompts and it's a bit more restricted it creates some fantastic images. I especially recommend changing it from 256x256 to 512x512. Leave it at 50 steps, that's all that's needed.
I think it is intentionally not showing faces is because people would use this program for some devious stuff. It's not really a great idea to give the world this power. Idk, I might be wrong, I might be confusing this program with another similar.
The melty images do feel like dreams. Its like when you try to focus on something in a dream and by the time you focus it has changed and you forgot what you were trying to focus on.
The man parachuting might work better as: "man parachuting with a cornish coastline background" or something like that. Using the word background is an easy way to force AIs into whatever place you want.
I'm an experimental electronic musician, and for someone like me this superior program is a godsend for any visual still image for tracks (while being "original" in a sense)
I also tried reading in a dream once but I couldn't stop myself from reading the same sentence over and over and then when I tried to stop reading I couldn't this made me panic i guess and i woke up soon after.
Dalle 2 has been amazing for me to play around with, it's realistic place interpretations are just mind blowing. It's scary how well it can place Shrek or waluigi in front of the Tyne bridge in Newcastle, given a good enough description! The fact that it can make a reasonable interpretation of the small English town I grew up in, and even a realistic looking street view of the village I live in now just blows my mind. Also when it comes to actual art, it's basically a profile picture creation tool. My friend wanted a pirate chef dragonborn character for D&D and it was amazing how usable the images it created were.
@@MrStn just searching things like, 'cute synthwave duck with a lil hat' produced some amazing results. And most of them are real cohesive images, not jumbled messes of lines that only come together to create an image.
I love these videos about developments in AI Personally I can't wait to have AI that's good enough to generate high quality images for my D&D campaigns, or possibly AI that can handle some programming tasks. It's mind-bending to think about where the technology could go if can match/exceed human reliability since it's already proved that it can sometimes match/exceed human performance.
Ok but like… am I the only one who doesn’t kind all the little missing features or weird textures? Like just knowing it can do this out of thin air on its own… and most times it WORKS… like if I put “fallout 3 with Jesse pinkman” it actually put fallout 3 with a character that looked close to Jesse pinkman from breaking bad 😂 and yeah the face wasn’t exactly there, but it was good enough that I, as well as anyone else would IMMEDIATELY see the image and tell what it was… which is impressive… either way this Ai and app is amazing, and the memes ALONE are worth. Shows what’s possible in the future.
Hey Phil, something I think you might be interested in testing is generating images with this tool and then scaling them down and then upscaling them. Could be interesting?
I wasn’t able to read newspapers, but the script of an invented play in my dreams. Was pretty tragic and funny, basically a love story between a knight and a mermaid. I even corrected the text… through eating pralines/chocolate and changing their order in the packaging… Pretty random but I had to think about that dream at 0:29
I have found out that the AI works really well with scary stuff because scary stuff benefits of the uncanny valley. Some of my favorite images were, when I tiped in "Flesh prison nightmare" or simular cursed word combiations.
You forgot the most commonly generated result of DALL-E Mini, "Too much traffic, please try again"
Ik right!
lovw that image
Just keep clicking run until it works. It took me 30 seconds
Just use the colab version lol
@@doggod9754 where can I find that?
feel like this has great potential to resurface in 15 years as a nostalgic meme about the awkward uncanny middle step for AI
I've been thinking the same thing. I predict that in 10-15 years this uncanny AI will become its own niche art form with a dedicated community of hipsters. It will be called AI-core
@@aurin_komak Clueless surely there aren't far more sophisticated AIs not available to the general public as we speak
@@ShadonicX7543 who the hell has nostalgia to tech they've never seen
this might be the top comment in 15 years
@@aurin_komak I already love the look of abstract AI images because they look like my dreams so when AI is disgustingly realistic and takes my job as an artist i hope they remember the uncanney valley era we were in
I hope they give him access to DALL-E 2
What are the differences?
@@vivosmartphone2280 2>1
@@vivosmartphone2280 it’s far better, can create photoreal images, or perfect renditions of the style you give it
@@vivosmartphone2280 ray tracing
@@vivosmartphone2280 16 x the detail
You should really talk about DALL-E 2. It's still in beta, but holy crap you need to see the progress they've made. It's honestly mind boggling.
I do believe there's a really long wait list, but it's worth checking out. A huge step over this DALL-E version. Open AI really seem to know what they're doing
Open AI? More like closed AI, I can't get in
@@itchylol742 LET ME INNNNNNN
it's actually amazing, I keep trying to stick funny characters into real life locations and it's terrifyingly good.
it even got a good render of my home town or the village I live in now, and all look accurate for the type of housing around here.
only problem for a lot of people is there are a lot of limitations on the new ones, the worst being 'star WARS'.
Its a lot of fun for waluigi pictures though.
the fact the ai in dall-e 2 knows how light should react through glass and liquid to create the correct caustics is insane
Open ai is a bad company. The only difference to dalle Mini is that opn ai has mor gpu and bigger data Sets that's all. The quality of the immaes Skalen in relation to training power and dataset size. Open ai does nothing special. Just look at Googles Text to immage ai it's xalled imagine. The quality is even better then dalle2. Because they have even bigger data Sets and even more gpu. At the moment ther is no real scientific progres in this area in ai. And the worst thing is that the big ones dosnt get release to the puplic. Dalle Mini is developed by a community and not by a bad company. And they release the source Code and the training Checkpoints. Because of this dalle Mini is better
For context, Dall E mini was NOT created by Open AI, the creators of original Dall e and Dall E 2
Why not
@@Chorweiler because it isn't
that explains why it is so bad compared to this other one
I find it works incredibly well when you give it an art style, in particular "photorealistic", "courtroom sketch" is incredible, "funko pop" of x is also great
Adding a ", high quality art" at the end of the prompt gave the best results for me.
Also some conceptual situations like a doll- or crochet-version of objects or people. Muppets look good and movies work well as templates, as well as "trail cam footage".
Sometime, the results can be extremely good
try the "hyper realistic" input or "3D Model" Input
also, "in the style of [artist] " or "by [artist]" gives amazing results.
No, nothing 'funkopop' is great :)
This AI has no chill. I asked it once "show me how you see us humans" and in response I got 1 image with one half showing a monkey and the other half a really dumb looking depiction of a human, almost as if it was mocking how humans look ...and the rest was all skulls D:
...stopped asking for pictures after that
That's some I Have No Mouth And I Must Scream stuff. 😅
💀
Thanks Mayonnaise Man, you saved the day. What would we do without your infinite wisdom?!
@@TackerTacker guess i activated ackhsually mode there, sorry about that one.
@@nonsensicalhumanoid 😜✌
I don't know why but the "[politician] in a bath of beans" concept just cracks me up
If we put adolf in there would the title be 'ziek hienz'
This ai is incredible, unfortunately it's gotten super popular lately, meaning it's unlikely to work as the traffic is too much for the servers. Hope they increase their servers or make their own website or something.
And the comments are full of 4chan edgelord subhumans
@@Jayfive276 Eh, at least they're anti-groomer subhumans.
@@blunderingfool talking about an average being in 4 chan you can never know.
@@KingLich451 Still preferable to groomers. Detransition accounts are absolutely harrowing and the people responsible just up and washing their hands of the matter need a floor to ceiling bath.
@@blunderingfool what the fuck are you on about? Are you the concept of whatsboutism that taken physical form?
My favorite Ive seen is someone posted a pic of Margaret Thatcher in a latex suit and the AI did it worryingly well
You should try and apply for Dall-E 2, I bet you could get a version to test if you provide how you're a tech youtuber and often talk about similar stuff.
> stepsister stuck in a washing machine
I see what you did there, Philip
So sneaky
Oh wow, these pictures look a lot like how things look at a glance when you take magic mushrooms. It's nuts how similar AI visualizations are to that.
From experience?
Not really
That's not even the most advanced model openai has. DALLE-2 is right now only available to researchers and let me tell you, it's something special from what I've seen on the internet. Altho the pics shown were probably curated.
yea very shroomy but also kinda ketamine in rhe dark vibes
Holy shit you’re right lmfao.
Like that one snapchat filter that makes faces everywhere
The most interesting thing about DALL-E Mini is that the model is completely different to DALL-E 2 and slightly inferior even from DALL-E 1. Plus there are no restrictions on people or faces that DALL-E 2 has so arguably has more use cases for the general public.
It's funny how this has become a lot of online people's first introduction to AI and in a way I'm glad it's this lesser model without restrictions. It gets people talking about the ethics of using real people in completely faked images while not having a high enough quality to be able to trick people. Sort of like a training ground for the general public on AI. Plus if everyone had access to DALL-E 2 I think people might have been underwhelmed by some of those restrictions anyway since DALL-E 2 is best at creating the sort of stuff that doesn't spread as quickly in "Meme Culture" circles.
2:02 recipe for cake in the middle 2nd row
1 (18.25-ounce) package chocolate cake mix
1 can prepared coconut-pecan frosting
3/4 cup vegetable oil
4 large eggs
1 cup semi-sweet chocolate chips
3/4 cup butter or margarine
1 2/3 cup granulated sugar
2 cups all-purpose flour
Don't forget garnishes such as:
Fish-shaped crackers
Fish-shaped candies
Fish-shaped solid waste
Fish-shaped dirt
Fish-shaped ethylbenzene
Pull-and-peel licorice
Fish-shaped volatile organic compounds and sediment-shaped sediment
Candy-coated peanut butter pieces (shaped like fish)
1 cup lemon juice
Alpha resins
Unsaturated polyester resin
Fiberglass surface resins and volatile malted milk impoundments
9 large egg yolks
12 medium geosynthetic membranes
1 cup granulated sugar
An entry called: "How to Kill Someone with Your Bare Hands"
2 cups rhubarb, sliced
2/3 cups granulated rhubarb
1 tbsp. all-purpose rhubarb
1 tsp. grated orange rhubarb
3 tbsp. rhubarb, on fire
1 large rhubarb
1 cross borehole electromagnetic imaging rhubarb
2 tbsp. rhubarb juice
Adjustable aluminum head positioner
Slaughter electric needle injector
Cordless electric needle injector
Injector needle driver
Injector needle gun
Cranial caps
@@2kliksphilip it's from the portal cake recipe. I just copied & pasted it lol
Lol
@@cowclucklater8448 misinformation 👍🏽
Past couple of days has made DALL-E mini barely usable, hope they get it fixed soon.
5:52 Those might actually be pretty accurate depictions of the queen celebrating her 200th birthday. We'll just have to wait about 100 years to see. 🤔
She turned 200 in the 9th century BC, though? I don't think you know your history well enough
Uhh ño
My favorite dall-e generation is margaret thatcher meeting satan, and it just drew her meeting herself.
What a time to be alive!
This is probably the best depiction of what a dream looks like
You're completely correct, the more details you give it the more it knows what to make specifically, if you give it something general and abstract it makes a wide variety of things all mixed together and thus looks worse. Based on interpretation it can create a wide variety of novel contextual images but you should be specific about the context if you want a clear conclusion
I just absolutely love your dry delivery of the individual prompts.
0:50 that's actually because it wasn't trained on faces, that's because the creators of it didn't want it to be used for malicious purposes
The Smaug as a beanie baby itself was awesome!
I put in Freddie Mercury drinking tea, and it seemed to grab information from live performance photos of Freddie, and slap a tea cup in his hand instead of a mic.
I was not disappointed, lol.
It also helps to specify the style you want it to output. I came up with some awesome results for "farewell, expressionism" and "spider, da vinci" or "dog warrior, pencil".
Improves the results by a lot!
AI generated faces truly a nightmare fuel
As far as I can tell they do it to avoid people making deepfakes
the strangest thing is that after generating these frightening faces, I had such a feeling in my eyes that everything is crooked like in these photographs.
Great idea to give it just an abstract term or emotion, and see how it visualizes that!
To be fair a tank with ONLY one wheel could technically be interpreted as just one wheel
i definitely needed to see these!
Truly this is one of the uploads of all time, riveting
Either you're really unoriginal or a bot
What I find interesting is just how dreamlike these images look. Its understandable, as its basically the same idea to dreams, you take a lot of inputs and reconfigure them to fit the currend demand. And as our sense of reality is turned off during dreaming, it gets very weird. It has also similaritys to how things look if you are on a mushroom trip. And it makes one wonder if its a similar mechanism, or if its completly diferent but arriving at the same result.
From an artists perspective, I would use this to help with just thinking of what to draw and using it as a sort of underpainting or thumbnail sketch to work off of.
Giving the prompt generator HIGHLY specific words like refractive, talking to, rolling down, also helps bring two objects to life within each other. My first two were "Wizard talking to Amish people" and "Hypersphere rolling down a mountain". I got some fascinating results, some were far less so than others such as "tornado wearing sunglasses" (none of them hit the mark, one of them was a cartoony alien head, but the actual tornados themselves were fairly interesting). Getting too specific however, would just end up with it focusing on something extremely easy, like "anomalous dark spots within a tesselated garden" (pardon spelling mistake) just go me this black weirdness.
Here was my foray into it seeing a random twitter link in duckduckgo that I mysteriously clicked on that was totally unrelated to my search, it was fun. (chronological order):
Amish | i.imgur.com/bhP5OOz.png
Hypersphere | i.imgur.com/QDl9CGf.png
Unseen truth | i.imgur.com/CXp1N6K.png (was curious to see if grammar would do anything, and then I saw the beaches and decided I wouldn't think further on it.)
Tornado | i.imgur.com/9NB8NLy.png
Room | i.imgur.com/R9fxwbE.png
Dark | i.imgur.com/iMaIMFQ.png
Coins | i.imgur.com/7wUfB2D.png
Texture | i.imgur.com/GdeJhyn.png
A-pylon-clipse | i.imgur.com/Zvi2Ayw.png
Fruit | i.imgur.com/8T21Svc.png
I can't wait to be accepted into the Dalle 2 list. I've been waiting since April. It's a massive improvement over Dalle 1
I love these types of things so when I saw it in ur vid I knew it would be worth my time. For whatever my mind could only think of "Kratos eating a cheeseburger next to the leaning tower of pisa" and to my surprise it did it. I have no clue how they do it but I love it.
Please make more of this. I loved it!
Also, in case you read this which I'm not expecting, try Latent diffusion. I believe you can Google "LAION 400M Google colab" and you should see it. Although you need to get more creative with the prompts and it's a bit more restricted it creates some fantastic images. I especially recommend changing it from 256x256 to 512x512. Leave it at 50 steps, that's all that's needed.
6:13 I know its a meme, but damn this one kinda hit something for me. Feels like a bad dream, stuck in a washing machine
Type that in "feels like a bad dream stuck in a washing machine"
This is really cool.
I don't understand why I like your content so much. But please keep going
Those video game recreations were spot on.
It's gets really fun when you enter a complex riddle
I think it is intentionally not showing faces is because people would use this program for some devious stuff. It's not really a great idea to give the world this power. Idk, I might be wrong, I might be confusing this program with another similar.
I am so excited for this technology. creating something new altogether is hard to wrap around my head. it's so cool!
The melty images do feel like dreams. Its like when you try to focus on something in a dream and by the time you focus it has changed and you forgot what you were trying to focus on.
"Too much traffic, please try again." Thanks for DDoS Attacking the side! ._. (irony alert.)
I feel like this would be good for story boarding for like movies and shows.
I need more of this
This really is one of THE videos you ever made
Sometime, the results can be extremely good
try the "hyper realistic" input or "3D Model" Input
The man parachuting might work better as: "man parachuting with a cornish coastline background" or something like that. Using the word background is an easy way to force AIs into whatever place you want.
brilliant work!
Let us all have a go, I urgently need an image of Kim Jong Un doing the ice bucket challenge!
This was just 11 months ago...
I love how you're going into the AI rabbithole and showing us stuff I've never heard of with great insight.
i've been having so much fun with this ai model. it reminds me a lot of some animations from the movie waking life
i appreciate showing this treasure to us at the cost of disabling their website 😄
That was beautiful
It also does the death metal album cover with differently good known pokemon really well.
It's amazing, I wonder how it will be in the future
I'm an experimental electronic musician, and for someone like me this superior program is a godsend for any visual still image for tracks (while being "original" in a sense)
Remind me to revisit this in 10 years
ok
Man, I fucking love DALL-E. Been in the queue for DALL-E 2 for a while now...
One word prompts are actually quite fun
I also tried reading in a dream once but I couldn't stop myself from reading the same sentence over and over and then when I tried to stop reading I couldn't this made me panic i guess and i woke up soon after.
This shit is so dope, I have been using it with my friend for the past week since we heard about it!!!
Thank you
The content you know you need
that train one is art XD
Boris Johnson eating Mango on the Moon is a particularly good one
Dalle 2 has been amazing for me to play around with, it's realistic place interpretations are just mind blowing. It's scary how well it can place Shrek or waluigi in front of the Tyne bridge in Newcastle, given a good enough description!
The fact that it can make a reasonable interpretation of the small English town I grew up in, and even a realistic looking street view of the village I live in now just blows my mind.
Also when it comes to actual art, it's basically a profile picture creation tool. My friend wanted a pirate chef dragonborn character for D&D and it was amazing how usable the images it created were.
Your third point here was one of the first things I thought of when I saw some of the results in this video.
@@MrStn just searching things like, 'cute synthwave duck with a lil hat' produced some amazing results.
And most of them are real cohesive images, not jumbled messes of lines that only come together to create an image.
Great video!!!!!!!!!
I love DALL-E mini
Will you do a video on Dalle 2? You should apply for the waitlist.
The demogorgon ones are my favourite
Always interesting to see you talk about AI
Time to revisit. Midjourney v5
DALL E is so interesting. i want to have a try, people are pounding the link at the moment.
Video is gold👌👌🏅🏅🏅
the current situation dall e mini has right now is exactly why dall e 2 has a waiting list
I went downt he exact same rabbit hole when someone first showed me the site. Now I'm comfortably addicted.
it does the hindenburg really well
I love these videos about developments in AI
Personally I can't wait to have AI that's good enough to generate high quality images for my D&D campaigns, or possibly AI that can handle some programming tasks. It's mind-bending to think about where the technology could go if can match/exceed human reliability since it's already proved that it can sometimes match/exceed human performance.
Part 2 please 👍
Ok but like… am I the only one who doesn’t kind all the little missing features or weird textures? Like just knowing it can do this out of thin air on its own… and most times it WORKS… like if I put “fallout 3 with Jesse pinkman” it actually put fallout 3 with a character that looked close to Jesse pinkman from breaking bad 😂 and yeah the face wasn’t exactly there, but it was good enough that I, as well as anyone else would IMMEDIATELY see the image and tell what it was… which is impressive… either way this Ai and app is amazing, and the memes ALONE are worth. Shows what’s possible in the future.
Something tells me Boris in a bath of beans is this guy's favourite 💀
The only NFT's I subscribe to
I done a 'yellow duck wearing sunglasses at a rave' and it was very convincing
lolol that is so amazingly aterrorizing
2:43 that middle one looks like a “my reaction to that information” meme lol
Hey Phil, something I think you might be interested in testing is generating images with this tool and then scaling them down and then upscaling them. Could be interesting?
I looked up, Serj Tankian jumping on a pogo stick about to land into a septic tank.
man-made horrors beyond my comprehension is cool
I wasn’t able to read newspapers, but the script of an invented play in my dreams. Was pretty tragic and funny, basically a love story between a knight and a mermaid.
I even corrected the text… through eating pralines/chocolate and changing their order in the packaging…
Pretty random but I had to think about that dream at 0:29
I have found out that the AI works really well with scary stuff because scary stuff benefits of the uncanny valley. Some of my favorite images were, when I tiped in "Flesh prison nightmare" or simular cursed word combiations.
Worth remembering that this is a greatly reduced model, hence hte coarseness. The full dalle2 or imagen are dramatically better.
I can very easily remember what a bike would look like, but I’d need time to figure out how to draw one because I’ve never done so.
3kliksphilip in a bath of beans
I did one with jackie chan eating a pizza, chuck Norris singing with rick astley and batman eating a cucumber
i wrote.. photorealistic cow tree..
human computer...
cat eating cheese (very weird lol)
a carpet made of egg..
Burning Asbestos has the uncanny resemblance to a Faith No More album cover.
Nostalgia