11 ms·

Here is my perspective on these kinds of images. This kind of 'picture' usually comes from speed-painters who incorporate techniques like photobashing. As in in

by pram 4y ago

Here is my perspective on these kinds of images. This kind of 'picture' usually comes from speed-painters who incorporate techniques like photobashing. As in integrating 3D models and RL photos into their composition, or just painting over a 3D picture entirely. It was already a genre that highly incorporated computer assisted methods. There is a lot of doom and glooming going around, but honestly the modern process of creating 'concept art' was already extremely commodified and efficient. These weren't exactly your idealized vision of some artisan craftsman laboring weeks over a picture, they churned this stuff out in a few hours (if that)

ricardobeat 4y ago
I think you're not grasping the magnitude of the change. Creating even an average quality speed-painting requires tremendous amount of expertise in drawing, painting techniques, composition, lighting, perspective. It takes years of training. These models let anyone achieve similar results in minutes. Without any prior learning. It is not even lowering the bar, it is literally dropping the bar to the ground. Besides, stable diffusion is able to generate not only painterly scenery, but also photographic images that are almost indistinguishable from actual pictures (certainly helped by the fact images have a heavy digital look in our era).
- shreyshnaccount 4y ago
  it just got faster. what's the drama about?
  - oefnak 4y ago
    I think it will be interesting to see how this comment has aged in five years.
  - dougmwne 4y ago
    I am talentless and untrained. Now with a combo of prompts and img2img, I can create awesome results on any topic and in any style that I have the rights to use. That’s a 0 to 1 moment. It didn’t get faster for me, it went from impossible to possible.
    - a_shovel 4y ago
      "Any style" seems like an enormous stretch. There definitely seems to be some styles that AI favors, ones which I've seen described as "clutter the frame so you don't notice the flaws". It struggles with simpler styles. I have yet to see a flat black & white image generated by an AI that looked even passable.
      - simonw 4y ago
        How about this one? https://static.simonwillison.net/static/2022/dall-e-pencil-squirrel.jpg https://static.simonwillison.net/static/2022/dall-e-pencil-s... From https://simonwillison.net/2022/Jun/23/dall-e/ https://simonwillison.net/2022/Jun/23/dall-e/ Have you tried DALL-E or Stable Diffusion yet? I bet you could generate a black and white image that met your standards for being impressive, if you spent a few minutes on it. You can try Stable Diffusion free here: https://beta.dreamstudio.ai/ https://beta.dreamstudio.ai/
        dougmwne 4y ago
        I considered qualifying “any” but decided it required no qualification. I don’t know how many examples are needed in the training data for a given artist to be able to reproduce that artist, but given how many obscure artists I have seen Dall-e and Stable Diffusion able to recognize, it must not require that many examples. And it’s still possible to fine tune a model with additional training if a new artist comes along or you want a bit more capability with a rare style. So yeah ANY style, I’m pretty sure of it.
        a_shovel 4y ago
        Nah, that's not it. I mean flat #000000/#ffffff. Google "stencil image" and you'll see what I mean. AI really doesn't handle styles with restrictions like that well. I tried the stable diffusion website with variations on "black and white silhouette stencil image of a cat". It kept wanting to give the cat colored eyes, or it used shading, or the cat didn't have a coherent anatomy, along with the typical AI art "duds" that aren't really anything at all. To be fair, I did get a couple of passable results when I replaced 'cat' with 'dog'. They were simple, but didn't have any obvious errors. To be fair in the other direction, replacing 'cat' with 'abacus' gave me an (admittedly pretty) grid of numbers and some chainmail, and 'helicopter' suggested a novel design where two helicopter bodies would be stacked vertically, connected by a vertical shaft through the rotor, and which turned into a palm tree trunk above the top unit. Once you get out of the sample data, it starts to fall apart. I feel like other people here are willing to forgive more errors than I am. They see an incoherent splotch in an image and assume more development can iron out all the problems, and I see a unavoidable artifact of the fact that these systems don't have a real understanding of what they're making.
  - flyinglizard 4y ago
    There’s nothing a modern computer does compared to 30 years ago other than being “faster”. It opens up everything.
    - pram 4y ago
      Technically true, but I think this is VASTLY understating what has become possible with your average PC over the past 30 years. Today I can get quick, effortless renders from Blender with a zillion available assets on the internet on my laptop. I can drop that directly into something like Clip Studio and paint right over it. In the 80s you needed an extraordinarily expensive workstation like the Quantel Paintbox to even do primitive Photoshop type stuff. If you wanted a 3D render you needed a whole farm of servers.
    - dougabug 4y ago
      That seems like an overreach. Thirty years ago, object recognition basically didn’t work, except in extremely simple cases. Something like semantic segmentation would have been way out of reach. Computers couldn’t play Go effectively against even modestly skilled human amateurs.
      - flyinglizard 4y ago
        I meant the very technical sense where you could take an object recognition algorithm and compile it to run on a 80386 and it would run fine although slow to the point of not being practical. Computers brought us more speed (and memory) to enable new classes of uses, but there’s not a single intrinsic operation a modern computer does which an old one can’t replicate. So quantity is indeed a quality.
- avidphantasm 4y ago
  But it didn't drop the bar on the ground, it raised a new bar. People without computer literacy and/or basic programming skills won't be able to pull this off. Even using Photoshop (which I believe does/will integrate this new technology) is not easy/possible for some who can actually draw. Plus, how many regular people have access to the machine with 12GB of VRAM?
  - coding123 4y ago
    It dropped the bar to the other side of the planet. There are so many people computer literate that can pull this off. You could pick 5 people off the street that can follow these instructions and 3 of them would. VS the old way, you would have to pull off 400 people off the street and you probably wouldn't get this result unless you got really lucky.
    - ipaddr 4y ago
      Pick 5 people at random you get one who doesn't know what a mouse is, 2 or 3 who can turn on the computer and maybe one who can get to the cli. Out of 400 people you will find more natural artists compared to someone who could install this even if they had the equipment
      - visarga 4y ago
        They can use it from their phone or tablet.
      - bee_rider 4y ago
        There are already web sites that will run the underlying models, requiring no installation. The neat new applications that have taken over this site for the last couple days sometimes require CLI steps to install because they are in active development and it can be easier to experiment with something local. I'm sure they'll either be moved online or wrapped in nice installers over the next couple weeks.
      - dougmwne 4y ago
        Let me update your heuristics on this. Computer mice are practically obsolete. People use cell phones, not computers. No one needs a cli to run stable diffusion because a mobile web interface was released on day one. 6.6 billion people have a smartphone which is 83% of the world’s population (including the infants). This is about the same number of people who are literate. 4 out of 5 people globally would be able to submit a stable diffusion prompt and view a result. Most would have no idea what the hell was going on or even why it was interesting.
- StrictDabbler 4y ago
  I agree. People aren't grasping the magnitude because they're thinking about jobs. Jobs a silly way to measure this. Jobs are temporary. Nobody worries about the mechanical stocking frame making socks anymore. This is more like the literacy/printing press transition. Used to be, people had to learn to memorize a lifetime of stories and lore. Now nobody learns to make a memory palace or form a mnemonic couplet. Why would you bother? You just write things down. Today, people learn to draw. In a generation, why would you bother? There will still be specialist jobs for people generating images, but instead of learning to make them up, the specialists will be very good at picking them, suggesting them, consuming them. Humans will be the managers and the editors, not the creators. The same thing will happen to other arts. First (and very easily) to music. Eventually, perhaps, to writing and whole movies... The only thing stopping that is that the models can't maintain a reality between frames. They can't make an arc. It's all dreamlike. If we find a way to nail object persistence it will be a singularity-level event. The moment you can say "make another version of this movie, but I want Edgar to be more sarcastic and Lisa should break up with him in the second act" we will close the feedback loop. It's a lot bigger than "lost jobs".
  - rvz 4y ago
    > It's a lot bigger than "lost jobs". I agree. It is more than just "lost jobs", like artist impressionists, court room sketch artists, etc. it is a complete dystopia and it doesn't help artists at all, but displaces them. At least the value of actual paintings will be more valuable that the abundance of this highly generated digital rubbish. So given that the technologists have so-called 'democratized' and cheapened digital art, I really can't wait until we get an open-source version of Copilot AI that would create full programs, apps, full stack websites with no-code so that we would be seeing very cheap Co-pilot AI shops in the south east of the world generating software that effectively eliminates the need for a senior full stack engineer. Easy cheap business solution for the majority of engineering managers on a tight budget who know they need to offshore tech jobs without the need for any skill as it is offloaded to cheap Copilot prompters. So we will have no problems with that and be happy with that dystopia. Wouldn't we?
    - 4y ago
- coffee_beqn 4y ago
  Speed painting also usually involves practicing certain scenes. This method anyone can use to create any new scene that they can imagine and the result looks quite good with some patience. Seems like some people are overly pessimistic but to me this seems like we’re on the cusp of something truly disruptive in the arts space. And it’s not NFTs. Remember that last year this would have sounded mostly like sci-fi unless you were following cutting edge research. In the realm or “real” art I’m actually very excited since I believe there are hundreds of very imaginative and patient people who just can’t paint well but will be able to create new art with tools like this. It can also synthesize new and alien things.
  - rvz 4y ago
    > This method anyone can use to create any new scene that they can imagine and the result looks quite good with some patience. Seems like some people are overly pessimistic but to me this seems like we’re on the cusp of something truly disruptive in the arts space. A race to the bottom and the cheapening of 'art' in general for the sake of replacing artists is a shame to see and nothing to celebrate. I was against both the gatekeeping of GPT-3 and DALL-E by Open 'faux' AI. But now it seems that every-time an open-source alternative or version was released into the wild, it seems that the uses become even more dystopian; especially with DeepFakes, fake news propaganda / articles and catfishing with generated hyperrealistic faces. > And it’s not NFTs. Remember that last year this would have sounded mostly like sci-fi unless you were following cutting edge research. Stable Diffusion is the reason why JPEG NFTs will always be worthless. Both of them will fuel JPEG NFT prices to the floor value of zero. But as NFT proponents cheered in believing that they will help artists, here we are seeing DALL-E 2 and Stable Diffusion fans screaming that it will help artists. No it will not. > In the realm or “real” art I’m actually very excited since I believe there are hundreds of very imaginative and patient people who just can’t paint well but will be able to create new art with tools like this. It can also synthesize new and alien things. This isn't the 'democratization of digital art', it is the complete devaluation and displacement of digital artists and it now makes 'real art paintings' much valuable. A dystopian creation.
    - dwohnitmok 4y ago
- wpietri 4y ago
  > it is literally dropping the bar to the ground I think you're directionally correct, but overstating the case in a few ways. One, as a not particularly visual person, even this example involves some skills of composition and perspective. If you asked me to do something practical, like creating an illustration to go at the top of a blog post, I would not do nearly as well as somebody with art skills, and I would take a lot longer. Two, this is the beginning. In the same way that digital artists took tools I could use and got really good at them, I expect the same will happen here. What will a good artist be able to do with a solid workflow and a few years of picking up tricks? Given the opaqueness and quirkiness of models, I expect a person who puts in the time, especially one with a strong command of art styles, composition, and the practical uses of visuals, will be able to run rings around me. Three, people are quite accepting of AI images right now, but they're novel and exciting and decontextualized from how we normally use images. That's a playing field that advantages the novice. But what happens once these images are no longer fun and novel, but boring and overdone? As we learn to discern novice-grade work from what real artists can do with AI assistance, I think our bar as image consumers will rise.
- Geee 4y ago
  Spending a lifetime learning img2img and using weeks to create a single artwork will always beat someone without experience who creates an artwork in an hour. There will be only a handful of people who will put in the time to become true masters of img2img. Everything comes always down to how much actual physical time someone is willing to put it. No matter how advanced the tools become, there's always a learning curve to mastery, which only a few people are willing or able to climb.

by pram 4y ago

ricardobeat 4y ago

I think you're not grasping the magnitude of the change. Creating even an average quality speed-painting requires tremendous amount of expertise in drawing, painting techniques, composition, lighting, perspective. It takes years of training. These models let anyone achieve similar results in minutes. Without any prior learning. It is not even lowering the bar, it is literally dropping the bar to the ground. Besides, stable diffusion is able to generate not only painterly scenery, but also photographic images that are almost indistinguishable from actual pictures (certainly helped by the fact images have a heavy digital look in our era).

shreyshnaccount 4y ago
it just got faster. what's the drama about?
- oefnak 4y ago
  I think it will be interesting to see how this comment has aged in five years.
- dougmwne 4y ago
  I am talentless and untrained. Now with a combo of prompts and img2img, I can create awesome results on any topic and in any style that I have the rights to use. That’s a 0 to 1 moment. It didn’t get faster for me, it went from impossible to possible.
  - a_shovel 4y ago
    "Any style" seems like an enormous stretch. There definitely seems to be some styles that AI favors, ones which I've seen described as "clutter the frame so you don't notice the flaws". It struggles with simpler styles. I have yet to see a flat black & white image generated by an AI that looked even passable.
    - simonw 4y ago
      How about this one? https://static.simonwillison.net/static/2022/dall-e-pencil-squirrel.jpg https://static.simonwillison.net/static/2022/dall-e-pencil-s... From https://simonwillison.net/2022/Jun/23/dall-e/ https://simonwillison.net/2022/Jun/23/dall-e/ Have you tried DALL-E or Stable Diffusion yet? I bet you could generate a black and white image that met your standards for being impressive, if you spent a few minutes on it. You can try Stable Diffusion free here: https://beta.dreamstudio.ai/ https://beta.dreamstudio.ai/
      - dougmwne 4y ago
        I considered qualifying “any” but decided it required no qualification. I don’t know how many examples are needed in the training data for a given artist to be able to reproduce that artist, but given how many obscure artists I have seen Dall-e and Stable Diffusion able to recognize, it must not require that many examples. And it’s still possible to fine tune a model with additional training if a new artist comes along or you want a bit more capability with a rare style. So yeah ANY style, I’m pretty sure of it.
      - a_shovel 4y ago
        Nah, that's not it. I mean flat #000000/#ffffff. Google "stencil image" and you'll see what I mean. AI really doesn't handle styles with restrictions like that well. I tried the stable diffusion website with variations on "black and white silhouette stencil image of a cat". It kept wanting to give the cat colored eyes, or it used shading, or the cat didn't have a coherent anatomy, along with the typical AI art "duds" that aren't really anything at all. To be fair, I did get a couple of passable results when I replaced 'cat' with 'dog'. They were simple, but didn't have any obvious errors. To be fair in the other direction, replacing 'cat' with 'abacus' gave me an (admittedly pretty) grid of numbers and some chainmail, and 'helicopter' suggested a novel design where two helicopter bodies would be stacked vertically, connected by a vertical shaft through the rotor, and which turned into a palm tree trunk above the top unit. Once you get out of the sample data, it starts to fall apart. I feel like other people here are willing to forgive more errors than I am. They see an incoherent splotch in an image and assume more development can iron out all the problems, and I see a unavoidable artifact of the fact that these systems don't have a real understanding of what they're making.
- flyinglizard 4y ago
  There’s nothing a modern computer does compared to 30 years ago other than being “faster”. It opens up everything.
  - pram 4y ago
    Technically true, but I think this is VASTLY understating what has become possible with your average PC over the past 30 years. Today I can get quick, effortless renders from Blender with a zillion available assets on the internet on my laptop. I can drop that directly into something like Clip Studio and paint right over it. In the 80s you needed an extraordinarily expensive workstation like the Quantel Paintbox to even do primitive Photoshop type stuff. If you wanted a 3D render you needed a whole farm of servers.
  - dougabug 4y ago
    That seems like an overreach. Thirty years ago, object recognition basically didn’t work, except in extremely simple cases. Something like semantic segmentation would have been way out of reach. Computers couldn’t play Go effectively against even modestly skilled human amateurs.
    - flyinglizard 4y ago
      I meant the very technical sense where you could take an object recognition algorithm and compile it to run on a 80386 and it would run fine although slow to the point of not being practical. Computers brought us more speed (and memory) to enable new classes of uses, but there’s not a single intrinsic operation a modern computer does which an old one can’t replicate. So quantity is indeed a quality.
avidphantasm 4y ago
But it didn't drop the bar on the ground, it raised a new bar. People without computer literacy and/or basic programming skills won't be able to pull this off. Even using Photoshop (which I believe does/will integrate this new technology) is not easy/possible for some who can actually draw. Plus, how many regular people have access to the machine with 12GB of VRAM?
- coding123 4y ago
  It dropped the bar to the other side of the planet. There are so many people computer literate that can pull this off. You could pick 5 people off the street that can follow these instructions and 3 of them would. VS the old way, you would have to pull off 400 people off the street and you probably wouldn't get this result unless you got really lucky.
  - ipaddr 4y ago
    Pick 5 people at random you get one who doesn't know what a mouse is, 2 or 3 who can turn on the computer and maybe one who can get to the cli. Out of 400 people you will find more natural artists compared to someone who could install this even if they had the equipment
    - visarga 4y ago
      They can use it from their phone or tablet.
    - bee_rider 4y ago
      There are already web sites that will run the underlying models, requiring no installation. The neat new applications that have taken over this site for the last couple days sometimes require CLI steps to install because they are in active development and it can be easier to experiment with something local. I'm sure they'll either be moved online or wrapped in nice installers over the next couple weeks.
    - dougmwne 4y ago
      Let me update your heuristics on this. Computer mice are practically obsolete. People use cell phones, not computers. No one needs a cli to run stable diffusion because a mobile web interface was released on day one. 6.6 billion people have a smartphone which is 83% of the world’s population (including the infants). This is about the same number of people who are literate. 4 out of 5 people globally would be able to submit a stable diffusion prompt and view a result. Most would have no idea what the hell was going on or even why it was interesting.
StrictDabbler 4y ago
I agree. People aren't grasping the magnitude because they're thinking about jobs. Jobs a silly way to measure this. Jobs are temporary. Nobody worries about the mechanical stocking frame making socks anymore. This is more like the literacy/printing press transition. Used to be, people had to learn to memorize a lifetime of stories and lore. Now nobody learns to make a memory palace or form a mnemonic couplet. Why would you bother? You just write things down. Today, people learn to draw. In a generation, why would you bother? There will still be specialist jobs for people generating images, but instead of learning to make them up, the specialists will be very good at picking them, suggesting them, consuming them. Humans will be the managers and the editors, not the creators. The same thing will happen to other arts. First (and very easily) to music. Eventually, perhaps, to writing and whole movies... The only thing stopping that is that the models can't maintain a reality between frames. They can't make an arc. It's all dreamlike. If we find a way to nail object persistence it will be a singularity-level event. The moment you can say "make another version of this movie, but I want Edgar to be more sarcastic and Lisa should break up with him in the second act" we will close the feedback loop. It's a lot bigger than "lost jobs".
- rvz 4y ago
  > It's a lot bigger than "lost jobs". I agree. It is more than just "lost jobs", like artist impressionists, court room sketch artists, etc. it is a complete dystopia and it doesn't help artists at all, but displaces them. At least the value of actual paintings will be more valuable that the abundance of this highly generated digital rubbish. So given that the technologists have so-called 'democratized' and cheapened digital art, I really can't wait until we get an open-source version of Copilot AI that would create full programs, apps, full stack websites with no-code so that we would be seeing very cheap Co-pilot AI shops in the south east of the world generating software that effectively eliminates the need for a senior full stack engineer. Easy cheap business solution for the majority of engineering managers on a tight budget who know they need to offshore tech jobs without the need for any skill as it is offloaded to cheap Copilot prompters. So we will have no problems with that and be happy with that dystopia. Wouldn't we?
  - 4y ago
coffee_beqn 4y ago
Speed painting also usually involves practicing certain scenes. This method anyone can use to create any new scene that they can imagine and the result looks quite good with some patience. Seems like some people are overly pessimistic but to me this seems like we’re on the cusp of something truly disruptive in the arts space. And it’s not NFTs. Remember that last year this would have sounded mostly like sci-fi unless you were following cutting edge research. In the realm or “real” art I’m actually very excited since I believe there are hundreds of very imaginative and patient people who just can’t paint well but will be able to create new art with tools like this. It can also synthesize new and alien things.
- rvz 4y ago
  > This method anyone can use to create any new scene that they can imagine and the result looks quite good with some patience. Seems like some people are overly pessimistic but to me this seems like we’re on the cusp of something truly disruptive in the arts space. A race to the bottom and the cheapening of 'art' in general for the sake of replacing artists is a shame to see and nothing to celebrate. I was against both the gatekeeping of GPT-3 and DALL-E by Open 'faux' AI. But now it seems that every-time an open-source alternative or version was released into the wild, it seems that the uses become even more dystopian; especially with DeepFakes, fake news propaganda / articles and catfishing with generated hyperrealistic faces. > And it’s not NFTs. Remember that last year this would have sounded mostly like sci-fi unless you were following cutting edge research. Stable Diffusion is the reason why JPEG NFTs will always be worthless. Both of them will fuel JPEG NFT prices to the floor value of zero. But as NFT proponents cheered in believing that they will help artists, here we are seeing DALL-E 2 and Stable Diffusion fans screaming that it will help artists. No it will not. > In the realm or “real” art I’m actually very excited since I believe there are hundreds of very imaginative and patient people who just can’t paint well but will be able to create new art with tools like this. It can also synthesize new and alien things. This isn't the 'democratization of digital art', it is the complete devaluation and displacement of digital artists and it now makes 'real art paintings' much valuable. A dystopian creation.
  - dwohnitmok 4y ago
wpietri 4y ago
> it is literally dropping the bar to the ground I think you're directionally correct, but overstating the case in a few ways. One, as a not particularly visual person, even this example involves some skills of composition and perspective. If you asked me to do something practical, like creating an illustration to go at the top of a blog post, I would not do nearly as well as somebody with art skills, and I would take a lot longer. Two, this is the beginning. In the same way that digital artists took tools I could use and got really good at them, I expect the same will happen here. What will a good artist be able to do with a solid workflow and a few years of picking up tricks? Given the opaqueness and quirkiness of models, I expect a person who puts in the time, especially one with a strong command of art styles, composition, and the practical uses of visuals, will be able to run rings around me. Three, people are quite accepting of AI images right now, but they're novel and exciting and decontextualized from how we normally use images. That's a playing field that advantages the novice. But what happens once these images are no longer fun and novel, but boring and overdone? As we learn to discern novice-grade work from what real artists can do with AI assistance, I think our bar as image consumers will rise.
Geee 4y ago
Spending a lifetime learning img2img and using weeks to create a single artwork will always beat someone without experience who creates an artwork in an hour. There will be only a handful of people who will put in the time to become true masters of img2img. Everything comes always down to how much actual physical time someone is willing to put it. No matter how advanced the tools become, there's always a learning curve to mastery, which only a few people are willing or able to climb.