I would LOVE to see what this AI produces when you expand the scene around famous portraits. What was the Mona Lisa wearing on her legs? Joggers? Jeans? What was Munch's Scream screaming at? What did the Laughing Cavalier find so darn funny? I will love and adore you forever if you make that video! 🤩
21:30 When he said it was Greg’n time and greged everywhere definitely my favourite part. Seriously though another banger and very creative vid from Greg!
Taking comforting images and generating discomforting/uncanny keywords outside could be interesting 🤔 E.g: a relaxing pool room, but the AI has to surround it with "thalassophbia" somehow
yesss imagine the monsters it'd create - it creates them without prompts anyway, so a prompt would make it just crack it's knuckles and go "alright kids, hold onto your hats"
11:05 is pretty spooky with how it looks like there was a violent murder scene on the left side to contrast the rest if the eerily peaceful liminal space. It really gives a sense of how casual and natural/normal the insanity is in this nearly infinite dimension, making it even more unsettling.
First found your channel back in April and have been a fan since. Perfect vibes, man. Oddly nostalgic really. It’s just like hanging out with childhood friends in the early 2000s just with new things to talk about lol keep up the great content, I’m loving it all.
15:25 I am also impressed by reflections. I see good reflections when using stable diffusion to generate a forest with a river through it - treelines appear properly shaped and upside-down in the reflection. Minor differences that wouldn't detract much from the experience of viewing a painting.
Love this! One tip - when you're making a selection for Generative Fill, include a bit of the original image so the AI has a little more to go on. In essence, select a smaller rectangle than the actual original image and then invert it for Generative Fill ;)
I love your AI series! You approach it with a healthy amount of awe and skepticism that I appreciate. Using it to expand backrooms images was a cool idea, I'm looking forward to seeing what else you use it for! Expanding creepy underwater photos could be interesting.
The image at 8:50 looks like it belongs in Resident Evil. There are definitely zombies hiding around those corners! I would love to see a video like this with liminal space and one with your own photos.
I just that discovered that you are also the TH-camr Gorg 🤣🤣 I guess two of my favorite TH-camrs are the same guy. How did I not realize this sooner???
I have a request, oh great and impressive Gregbrodudeman. Sleep story 😃 you've got one of those voices that is just nice to listen to. You're not screaming or howling. You're chill and I respect that. Like when you read us your diary, read us something else chill. A bedtime story in your voice......hell yeah😉
OG oggy and the cockroaches episodes has these part where they show a comically huge house interior with almost nothing but a few furnitures and appliances that shit gives the vibe of liminal space kinda
You have a beautifully dry and sardonic sense of humour. I suspect that a mistake was make somewhere in your manufacture and you should have been born British. I'm a Brit, consider my statement the highest praise. :)
The worst thing about the shallow pool backroom is that over time, you'll begin to have problems like trench foot and basically slowly fall apart because your body isn't designed to be wet all the time.
20:40 But you know what _is_ in your original image? ambient light faintly influenced by the presence of a wooden floor. Audio equipment associated with a profession, art taste associated with a culture, and each of those associated with floor types. It will master all of these correlations long before it masters fingers. Probably because this affects more of the image, so it has more importance during training. But also probably because nearly every indoor image in the training data trains the floor-prediction ability, whereas I'm pretty sure the hands problem cannot be solved by throwing more pictures of hands at the problem. It's already seen several million hands with 5 fingers, so if it generates one with 6 fingers, it has other issues. But ALSO, maybe it guessed the floor correctly because images from your streams were in the training data.
Nobody will talk about that painting in your generated room that looks like a portrait of a famous painter that started a world war but with red devil horns?
I would love more generative ai stuff, but a suggestion is to make sure that when you use the select tool, to make sure you dont accidentally also overlap the white background, a few times you did it and you can see how that affected the ai in its generation, trying to interpret a bizarre white outline
I have 2 cool concepts to try, one is that you put a bunch of pictures in one image, and have the ai try to generate inbetween, and the other idea is to generate an image, than highlight a section and try to get it to generate based on a portion of an image it generated.
for alot of these you clould have just magic wanded the whitespace around the image and then generated, saving you from having to manually select and then invert. tho you may have had to make the magic wand increase the selection by a couple pixels, idk how good photoshops magic wand is probably wouldn't have but I guess it's easier for the layman if you don't have to mess with settings and to have one thing you use every time, but magic wand would've been my instinct for this kind of thing
The end is near picture is actually a real location. It's the inside if an old border's bookstore either in Chicago or California. I'm not sure which location but I know it is a real place. Just a fact I wanted to point out.
I would love if you ai generate like famous liminal space images. Or like even thalassophobia or submechanophobia would be cool too. Either way I would love to see more.
Ideas for AI FIll-In - famous landmarks - famous photos - famous paintings - animals. - biology - words crop out portions of well known things and see if the AI can mirror it correctly or fill out the rest of it in a believable way. (perhaps even creating a new art piece as it copies the original style and tries to create around it) Like if you took the Eiffel Tower, cropped out the top of it, or did only one half of it, would it be able to "build" the eiffel tower based only on that limited reference? or make something entirely new? for animals take something like a dog's front half, and see if it will give the dog appropriate back legs. biology would be interesting because if you gave it part of a picture of a human skeleton, would it be able to create a full skeleton? and for some bonus stuff , can you go further down the spiral by generating an image, take the AI part, have it generate from only THAT AI piece, and see what it fills in from the other half? words. if you write something, crop the bottom/top half of all the letters out, could the AI fill in the gaps and rebuild the letters to make the words legible again? or would it create an entirely new alphabet?
I'd love (and hate) a generative fill thalassophobia video. It'd be interesting to see what manner of spooky stuff AI could think to put in the ocean.
This is a great idea!
Yes!!
Came here to say just this!
Oh my gosh, yes!
He’s already done something similar but it would be great if he did it a 2nd time
I would LOVE to see what this AI produces when you expand the scene around famous portraits. What was the Mona Lisa wearing on her legs? Joggers? Jeans? What was Munch's Scream screaming at? What did the Laughing Cavalier find so darn funny? I will love and adore you forever if you make that video! 🤩
Didn’t expect to get jump scared in the final AI generation set.
A Greg day is a good day
Agree!!!
Yes
YES ❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤ ❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤ ❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤ ❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤ ❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤ ❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤ ❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤ ❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤❤
The jump scare of the woman generating out of thin air is gonna give me nightmares
@Snappy33-is-cool. Hi!
21:30 When he said it was Greg’n time and greged everywhere definitely my favourite part.
Seriously though another banger and very creative vid from Greg!
Taking comforting images and generating discomforting/uncanny keywords outside could be interesting 🤔
E.g: a relaxing pool room, but the AI has to surround it with "thalassophbia" somehow
yesss imagine the monsters it'd create - it creates them without prompts anyway, so a prompt would make it just crack it's knuckles and go "alright kids, hold onto your hats"
11:05 is pretty spooky with how it looks like there was a violent murder scene on the left side to contrast the rest if the eerily peaceful liminal space. It really gives a sense of how casual and natural/normal the insanity is in this nearly infinite dimension, making it even more unsettling.
First found your channel back in April and have been a fan since. Perfect vibes, man. Oddly nostalgic really. It’s just like hanging out with childhood friends in the early 2000s just with new things to talk about lol keep up the great content, I’m loving it all.
15:25 I am also impressed by reflections. I see good reflections when using stable diffusion to generate a forest with a river through it - treelines appear properly shaped and upside-down in the reflection. Minor differences that wouldn't detract much from the experience of viewing a painting.
0:27- AHHHH !
Love this! One tip - when you're making a selection for Generative Fill, include a bit of the original image so the AI has a little more to go on. In essence, select a smaller rectangle than the actual original image and then invert it for Generative Fill ;)
I love your AI series! You approach it with a healthy amount of awe and skepticism that I appreciate. Using it to expand backrooms images was a cool idea, I'm looking forward to seeing what else you use it for! Expanding creepy underwater photos could be interesting.
The image at 8:50 looks like it belongs in Resident Evil. There are definitely zombies hiding around those corners!
I would love to see a video like this with liminal space and one with your own photos.
I just that discovered that you are also the TH-camr Gorg 🤣🤣 I guess two of my favorite TH-camrs are the same guy. How did I not realize this sooner???
I loooooooove the last one. It was crazy to see how far it could go. Though your other hand was an abomination.
16:41 had me laughing way harder than it should've lmfao
you know it's a good day when greg posts
This concept was awesome! 🤩 I also gotta admit that I am quite jealous of your eyelashes
@Snappy33-is-cool. Hello?
I love the honesty about being busy not wanting to post
This is a whole lot of FUN! I love seeing what AI is capable of! And with you as a commentator - its TH-cam gold!!
I have a request, oh great and impressive Gregbrodudeman. Sleep story 😃 you've got one of those voices that is just nice to listen to. You're not screaming or howling. You're chill and I respect that. Like when you read us your diary, read us something else chill. A bedtime story in your voice......hell yeah😉
Such an amazing video! You had me laughing so hard 3:30 ❤
20k subs in a few weeks.
Love you Greg.
Real friends understand you cannot pour from an empty cup, we will always wait for your cup to be full.
Man, you totally got what i would like to see in these pics too. Id Love to see more
I love that it gave you a co-host who looks as lost as the AI feels... And nightmare fuel demon 😂
Thank you brother this was sick!
You seem to "get" the aesthetic of the Backrooms well enough that I bet you could start making your own Backrooms videos.
this was really fun! i loved watching you mess with it and get impressed and breaking down what it did
11:30 reminded me of BSG hangar bay!
Thalassaphobia version of this would be super cool!
Please do more of these! The generative fill videos are always so interesting to me
Dude im shocked you didn't touch on the submersible imploding, would love to hear your thoughts cuz that's some thalassaphobia right there
It's okay Greg, I was busy last week too, so I didn't have time to watch a video anyways
Oh perfect, glad I didn’t waste my time!
This was super cool! I'd love more of these!
If you do go and take pictures you should take a wide shot and a zoomed shot then use AI on the zoomed one to compare and see how accurate it is
That disembodied head at the end was pretty great
OG oggy and the cockroaches episodes has these part where they show a comically huge house interior with almost nothing but a few furnitures and appliances that shit gives the vibe of liminal space kinda
You have a beautifully dry and sardonic sense of humour. I suspect that a mistake was make somewhere in your manufacture and you should have been born British. I'm a Brit, consider my statement the highest praise. :)
THIS WAS SO COOL! I think I need more of this!
A.I. - "Quit making me add nudes."
Greg - "Bruh? It's just a hallway you perv."
First of this kind of video I've seen with this generative AI with photoshop!!
I would love to see how much more you can do with it!!
This was really fun to watch! I would totally be freaked out at the random woman appearing in the room though... o_o
Your videos put me in such a good mood!
16:41 XDDDDDDDDDDD😂😂😂😂
The worst thing about the shallow pool backroom is that over time, you'll begin to have problems like trench foot and basically slowly fall apart because your body isn't designed to be wet all the time.
Photoshop and AI generation on the pictures was a neat idea Greg.
20:40 But you know what _is_ in your original image?
ambient light faintly influenced by the presence of a wooden floor. Audio equipment associated with a profession, art taste associated with a culture, and each of those associated with floor types. It will master all of these correlations long before it masters fingers.
Probably because this affects more of the image, so it has more importance during training.
But also probably because nearly every indoor image in the training data trains the floor-prediction ability, whereas I'm pretty sure the hands problem cannot be solved by throwing more pictures of hands at the problem. It's already seen several million hands with 5 fingers, so if it generates one with 6 fingers, it has other issues.
But ALSO, maybe it guessed the floor correctly because images from your streams were in the training data.
Came for the Thallasophobia, staying for the BroDude.
Real
love your videos❤
Amazing video, my dude ✌🏻😁 As Always haha
You could do this, but then select things you don't like after it generates, and then generate new bits in those areas.
love ur vids greg :)
That mcdonalds in the backrooms looks like ive been there before.
i died with the clean it up jump scare lol
The back rooms and ai. Bro just made the most trending idea possible.
The clean it up sent me lol
Greg time
3:23 oh yeaaah, blame it on the ai. Not you being sussy at all, right. Right? 😀
trippy...
Christ that last one was hella scary 😅
Super cool Idea for more videos! Do some underwater scary shit with this technology! :P
I literally can't wait for u to hit 100k
Ai backrooms stuff is always awesome
Nobody will talk about that painting in your generated room that looks like a portrait of a famous painter that started a world war but with red devil horns?
we all know the next step to this.. is seeing if it works with Thalasso stuff XD
I would love more generative ai stuff, but a suggestion is to make sure that when you use the select tool, to make sure you dont accidentally also overlap the white background, a few times you did it and you can see how that affected the ai in its generation, trying to interpret a bizarre white outline
I have 2 cool concepts to try, one is that you put a bunch of pictures in one image, and have the ai try to generate inbetween, and the other idea is to generate an image, than highlight a section and try to get it to generate based on a portion of an image it generated.
for alot of these you clould have just magic wanded the whitespace around the image and then generated, saving you from having to manually select and then invert.
tho you may have had to make the magic wand increase the selection by a couple pixels, idk how good photoshops magic wand is probably wouldn't have but I guess it's easier for the layman if you don't have to mess with settings and to have one thing you use every time, but magic wand would've been my instinct for this kind of thing
Love you brother
20:55 I might be wrong, but I swear that looks like Jeanie from MXR's channel
The end is near picture is actually a real location. It's the inside if an old border's bookstore either in Chicago or California. I'm not sure which location but I know it is a real place. Just a fact I wanted to point out.
Yo you don't have to do only scary stuff it's real nice to have varying content types
I'm pretty sure these are just his interests. He did have some non-scary stuff with the reading of his old journal 😂 I enjoyed those.
“Clean it up” had me dying
The last generative images were interesting to see, u really could just use this ai to help build room ideas
I would love if you ai generate like famous liminal space images. Or like even thalassophobia or submechanophobia would be cool too. Either way I would love to see more.
more of this 🤌🤌🤌🤌
I like this style of video
Feed ur diary into chat gpt and have it continue the story of greg and jessica
OH MY GOSH YEEEES!
200IQ right there also try a ufo thats in the sky
I only started watching your channel in the last month and I'm even upset that you didn't post last weekend a year ago.
Idk if pure liminal space images would generate something interesting but if you do a little test and see that it’s worth it, I’d love a video on it.
Huh interesting
This video was definitely cool and interesting
If you select areas you don't like in an image and generate in them, over time you should get a better and better image.
I wanna see more AI generated Squidward pictures.
19:48 bro AI added austrian painter picture to your wall💀💀💀
honestly, i’d be up for another AI video. like take well known places around the world and AI generate around them or something idk
when you gonna show us the frontrooms??
@19:34 bro has an AI generated Hitler poster 😂
You should try this on a skatepark
4:40 lol, it looks not bad
lol the little comments in the corner of the vid
The scariest thing is how criminally under-liked this video is.
do more of this greg part 2
gregbrodudeman
19:59 did the AI put a painting of Hi7ler on the wall? lol wtf
''what is this sleep paralyses demon in my corner''💀
How dare you be busy last week. Dude post when you want! Quality not quantity💕
Ideas for AI FIll-In
- famous landmarks
- famous photos
- famous paintings
- animals.
- biology
- words
crop out portions of well known things and see if the AI can mirror it correctly or fill out the rest of it in a believable way. (perhaps even creating a new art piece as it copies the original style and tries to create around it)
Like if you took the Eiffel Tower, cropped out the top of it, or did only one half of it, would it be able to "build" the eiffel tower based only on that limited reference? or make something entirely new?
for animals take something like a dog's front half, and see if it will give the dog appropriate back legs.
biology would be interesting because if you gave it part of a picture of a human skeleton, would it be able to create a full skeleton?
and for some bonus stuff , can you go further down the spiral by generating an image, take the AI part, have it generate from only THAT AI piece, and see what it fills in from the other half?
words. if you write something, crop the bottom/top half of all the letters out, could the AI fill in the gaps and rebuild the letters to make the words legible again? or would it create an entirely new alphabet?
That girl in the final ai set of your room looks like Jeannie from the channel MXRPlays