[HN Gopher] Using Stable Diffusion's img2img on some old Sierra ...
       ___________________________________________________________________
        
       Using Stable Diffusion's img2img on some old Sierra titles
        
       Author : Karuma
       Score  : 132 points
       Date   : 2022-09-05 17:24 UTC (5 hours ago)
        
 (HTM) web link (sciprogramming.com)
 (TXT) w3m dump (sciprogramming.com)
        
       | yummybear wrote:
       | Awesome, but it's totally missing the atmosphere in the originals
       | imo
        
       | iforgotpassword wrote:
       | I found the intro of one of the kings quest games run through SD
       | a few days ago, but can't seem to find it now. It wasn't that
       | impressive since I guess there wasn't much fine tuning going on,
       | but I liked the general idea. It had a few funny hiccups, but I'd
       | have expected much more erratic behavior, because there is no
       | information shared between frames (I guess). But maybe this
       | concept can be improved upon?
        
       | vslira wrote:
       | Is there any service offering user-friendly access to open source
       | DL models? Paid is fine.
        
         | fassssst wrote:
         | Midjourney uses Stable Diffusion now too in their beta version.
        
         | adammarples wrote:
         | The stable diffusion repo itself is free, and you can get the
         | model from hugging face, also for free. Runs on my 1070 easily.
        
         | fragmede wrote:
         | http://beta.dreamstudio.ai is the official paid Stable
         | Diffusion webUI.
        
           | Silverback_VII wrote:
           | I liked it until I noticed that it costs me $10 a day. I'm
           | also not sure if they support img2img. I have used
           | huggingface and then replicate for it(NSFW filter is very
           | annoying).
        
             | skybrian wrote:
             | Unless you're looking at something else, it's not per day,
             | it's a credit system that amounts to about 1 cent per
             | image. How many images do you plan to generate a day?
        
               | Silverback_VII wrote:
               | yes, it's per image but for some reason I easily rushed
               | through the 1000 generations in one day.
        
               | MattRix wrote:
               | Well you've gotta either run it locally or pay for it.
               | The GPUs this stuff runs on are too expensive to offer
               | unlimited use for free.
        
               | Silverback_VII wrote:
               | I'm not against paying but it should be somewhat
               | raisonnable. to put things into perspective: Colab pro
               | only costs about $10/month and you will probably be able
               | to generate at the same speed.
        
       | z9znz wrote:
       | The before and after are great.
       | 
       | There is an art to conveying a feeling with limited resources.
       | That's what makes early computer game images (the good ones...
       | because there were plenty of bad ones) so special.
       | 
       | The same could be said even more strongly for words. I'm not a
       | writer and I don't remember everything, but someone famous once
       | said something about eliminating everything non-essential from
       | writing to make it better. That's what makes a writer really
       | great.
       | 
       | Even so, the AI-upscaled versions of the original art are
       | impressive.
       | 
       | I said it before on this topic, and I'll repeat it. We will
       | someday (soonish) have games where the art is generated in real
       | time, unique for each player, based on good inputs. And it will
       | be awesome. Every play and every experience will be relatively
       | unique, but most or all of the plays will be excellent. That's an
       | exciting prospect.
        
       | selimnairb wrote:
       | Yeah, it's amazing that things like Stable Diffusion can make
       | coherent looking images. But why? Why convert iconic EGA and VGA
       | *art* into algorithmic representations? To me, it doesn't improve
       | on the originals in any way. I would never play a game with the
       | graphics replaced by these. To me, it is the constraints of older
       | technologies that led artists to create imaginative works that
       | transcended the materials available (big pixels with a limited
       | palette) and helped create worlds where your own imagination
       | could run wild. Replacing these images with photo realistic
       | Thomas Kinkade-esque treacle is an abomination.
        
         | MattRix wrote:
         | Why do people always ask "why?" on these things? Not everything
         | has to be "useful" or have commercial value. The original games
         | aren't going anywhere. People just enjoy doing this kind of
         | stuff, pushing the technology to make interesting things.
        
           | selimnairb wrote:
           | I'm not criticizing it because it's not useful. I'm
           | criticizing it because it is not art, and it is anti-
           | humanist.
        
             | ben_w wrote:
             | """At the 1983 Academy Awards, Oscar voters declined to
             | nominate [Tron's] pioneering special effects. Lisberger
             | said it was because they felt using computers as an
             | animation tool was cheating.""" -
             | https://www.moviefone.com/2017/07/08/19-things-you-never-
             | kne...
        
             | selimnairb wrote:
             | Oh, and the pictures look like shit.
        
       | decibe1 wrote:
       | This is really timely - I just started to replay some of my
       | favorite Sierra titles "Hero's Quest".
       | 
       | 1 and 2 (unofficially) have VGA ports that updated the graphics.
       | So, I may have to run some screen shots through to get an even
       | newer backdrop.
        
       | terrio wrote:
       | Awesome!
       | 
       | I thought about this in a slightly different way: D2 remastered
       | has a great switch back feature.
       | 
       | You could easily train a network with tons and tons of D2 old vs.
       | New and just use it to Auto upscale /reimagine of old D2.
       | 
       | Image rendering a old D2 video into 4k D2 remastered style.
        
       | davesque wrote:
       | Going to have to be the naysayer here. First, I'll say the simple
       | fact that Stable Diffusion produces anything coherent is
       | incredible. I'm blown away by the tech. However, my honest
       | opinion of the results showcased in this article is not positive.
       | Many of the result images contain bizarre distortions or
       | dreamlike artifacts that severely disrupt the flow of the image.
       | Especially the first one. It's clear that there's something like
       | a person with long hair standing in front of a peak shrouded in
       | clouds. But it only takes one or two moments to see the flaws in
       | the output. I wonder if hyper parameter tuning would help.
       | 
       | Again, I think the results are impressive in their own right. But
       | they seem impractical on account of the flaws in the details.
        
         | sportslife wrote:
         | The inconsistent and often incoherent shadows and lighting
         | effects are immediately annoying in all generated images if
         | you've spent any time working with visual media.
        
         | jamiek88 wrote:
         | Weird until I read your comment I was blown away. Then I had
         | another proper look at the first image and in many ways I had
         | to turn off my brains amazing 'upscaling' ability.
         | 
         | My brain had upscaled that human like blob to a woman spinning
         | around with a sword so her hair covered her face.
         | 
         | Looking closely. None of that is there really, just a
         | suggestion of it. And that is enough.
         | 
         | The more I learn about vision and sight the less I'm sure that
         | we see reality.
        
           | LocalH wrote:
           | We see reality in some sense - light of various wavelenghts
           | reflects off physical objects and enters our eye. However,
           | our processing and subjective interpretation of the input is
           | what can be more subjective, because every person's brain is
           | going to process such items differently based upon experience
           | (especially early life experience, when our brains are the
           | most plastic). Also, some people have sensory differences
           | (such as color blindness) that can influence the processing
           | of the light that enters our eye.
           | 
           | Objective reality exists, for some definition of "exists" -
           | there is physical matter present, with properties enabling
           | some or all wavelengths of light to reflect (and similar for
           | other senses like hearing). However, if we viewed reality
           | devoid of the subjective processing, we'd "see" everything,
           | but key existential concepts such as object permanence would
           | not be possible, as that requires our brain be able to
           | process and recognize an object in order to identify what the
           | object is in the first place, to even be able to remember
           | what it is. Not entirely unlike the iterative process of
           | modern machine learning.
        
           | 5d8767c68926 wrote:
           | >Looking closely. None of that is there really, just a
           | suggestion of it.
           | 
           | In contrast to the photorealistic pixel art?
        
         | ziml77 wrote:
         | I think the weirdness is most obvious in the cliff overlooking
         | a beach and in the woman's impossible object hand.
         | 
         | All of the output here is cool and impressive for sure, but
         | good art it is not.
        
         | idiotsecant wrote:
         | It's interesting to see how much this has improved in the space
         | of _weeks_ , imagine what this will look like in a year, in 10
         | years, in 30 years.
         | 
         | This is the beginning of something pretty big, I think.
        
         | fezfight wrote:
         | Compared to low res Sierra graphics, what's been shown here is
         | nothing short of astounding, if you ask me. Look at the skull!
         | It's a masterpiece.
        
         | AaronFriel wrote:
         | Are you naysaying Stable Diffusion or the idea of generative
         | models for art in general?
         | 
         | It's hard to look at what's happened in the last few months and
         | not think of it as akin to the invention of the steam engine,
         | but for art.
         | 
         | It's not perfect, as early machines had many flaws, were wildly
         | inefficient, produced irregular output. But the innovation that
         | followed created the industrial revolution.
        
         | jgg7 wrote:
         | What's missing is the charm. In general, AI-generated visual
         | content tends to have a very sterile look to it. The only image
         | in said thread that looks even remotely decent to me is the one
         | of a canyon/bridge. The rest are what I'd classify as voids.
        
           | Silverback_VII wrote:
           | >AI-generated visual content tends to have a very sterile
           | look to it
           | 
           | I think it's because the prompts are often sterile as well.
           | you have to add stuff like "matte painting" and "dream" just
           | like in the first example in dreamstudio."very detailed
           | landscape with xy" also works fine. Avoid prompts like
           | "digital art" or "render".
        
       | holoduke wrote:
       | Wonder if one day we can do realtime video conversion with this
       | tool on a phone using the camera. We be mind blowing to see it
       | live in action and changing your backyard in any style you like
       | to see. Maybe even with VR. Ultra tripping experience. I guess we
       | have to wait 5 to 10 years
        
         | ben_w wrote:
         | Something like this, but with an automated flow?:
         | https://twitter.com/karenxcheng/status/1564635828436885504
        
       ___________________________________________________________________
       (page generated 2022-09-05 23:01 UTC)