|
example: 
|
#
?
Sep 12, 2022 12:31
|
|
|
|
| # ? May 30, 2024 10:10 |
|
|
Someone should make a game where the AI generates all the assets on the fly. like that 3d shooter compressed into a megabyte
|
|
#
?
Sep 12, 2022 12:34
|
|
|
BARONS CYBER SKULL posted:example: Hmmm wondering if SD can generate convincing maps for tabletop gaming (over world, dungeon). Gonna try. Or maybe individual tiles. It probably won't be more efficient than the process my girlfriend has but can't hurt to try. She's been generating various AI and similar tools on her world building already anyway.
|
|
#
?
Sep 12, 2022 12:50
|
|
|
Rutibex posted:Someone should make a game where the AI generates all the assets on the fly. like that 3d shooter compressed into a megabyte the AI itself takes several gigabytes, though there are games like kkrieger that uses procedural generation, but it's not AI just hand tuned pattern generators
|
|
#
?
Sep 12, 2022 13:03
|
|
|
Lord Stimperor posted:Hmmm wondering if SD can generate convincing maps for tabletop gaming (over world, dungeon). Gonna try. Or maybe individual tiles. isometric [whatever] ruins on a colorful ocean, unreal engine, fantasy art by greg, loish, rhads, ferdinand knab, makoto shinkai, lois van baarle, ilya kuvshinov, rossdraws, tom bagshaw, global illumination, radiant light, highly detailed and intricate environment, trending onstudio ghibli, octane render, 8k  old prompt i had saved
|
|
#
?
Sep 12, 2022 13:22
|
|
|
Rutibex posted:Someone should make a game where the AI generates all the assets on the fly. like that 3d shooter compressed into a megabyte Hyper Compression with fractals already got this on a floppy 96kb it actually looks really loving good. https://www.youtube.com/watch?v=bD1wWY1YD-M e: AI made game from last year AI - GTA Bridge https://www.youtube.com/watch?v=udPY5rQVoW0 pixaal fucked around with this message at 13:57 on Sep 12, 2022 |
|
#
?
Sep 12, 2022 13:41
|
|
|
Lord Stimperor posted:Hmmm wondering if SD can generate convincing maps for tabletop gaming (over world, dungeon). Gonna try. Or maybe individual tiles. https://watabou.github.io/city-gene...ates=-1&sea=0.8 
|
|
#
?
Sep 12, 2022 14:09
|
|
|
ymgve posted:the AI itself takes several gigabytes, though Yeah but modern games are like 100gb. it could still be a savings
|
|
#
?
Sep 12, 2022 15:31
|
|
|
VectorSigma posted:edit: can't forget the history lesson. today we will learn about the lesser-known "monster truck crusade" this owns
|
|
#
?
Sep 12, 2022 15:48
|
|
|
i could easily imagine a year or two from now being able to link a game to a fine-tuned model hosted online and use it to generate all of the graphics. maybe not real-time animations or 3d, but for a 2d, strategy, or weird/casual game i think we'll see some absolutely insane stuff come out pretty quick
|
|
#
?
Sep 12, 2022 15:55
|
|
|
AARD VARKMAN posted:i could easily imagine a year or two from now being able to link a game to a fine-tuned model hosted online and use it to generate all of the graphics. I think you already can use it to generate assets for certain 2d games, with a little bit of tweaking. In a few years, you can probably use it for 3d graphics. It wouldn't surprise me if this is greatly improved upon in the future https://www.youtube.com/watch?v=_Y1-KlTEmwk. I think the era of lone creators being able to put out some really amazing stuff without having an insane work schedule is approaching.
|
|
#
?
Sep 12, 2022 16:07
|
|
|
IShallRiseAgain posted:I think you already can use it to generate assets for certain 2d games, with a little bit of tweaking. In a few years, you can probably use it for 3d graphics. It wouldn't surprise me if this is greatly improved upon in the future https://www.youtube.com/watch?v=_Y1-KlTEmwk. I hadn't seen any of that animated outpainting in that video from Google and it makes me feel like my dream of a planet-scale GTA (using streetview in its training) may actually come true some day.
|
|
#
?
Sep 12, 2022 16:25
|
|
|
AARD VARKMAN posted:I hadn't seen any of that animated outpainting in that video from Google and it makes me feel like my dream of a planet-scale GTA (using streetview in its training) may actually come true some day. I mean flight simulator already did a planetwide map using street view + satellite view. I don't believe AI was used to implement that though.
|
|
#
?
Sep 12, 2022 16:32
|
|
|
Objective Action posted:
I still want to shoot him. You rule.  IShallRiseAgain posted:I mean flight simulator already did a planetwide map using street view + satellite view. I don't believe AI was used to implement that though. MSFS uses AI tools to generate, texture, and place 3D buildings on the terrain mesh, with the data derived from 2D overheads. There is also hand-crafted scenery (a lot of cities and airports and such,) but the algorithm will generate passable scenery globally. It’s an impressive thing, even if the launcher/patcher is a war crime, and it requires ludicrous install sizes and bandwidth to function properly.
|
|
#
?
Sep 12, 2022 16:46
|
|
|
Rahu posted:I just tried using stable diffusion to make some textures for some light graphic design work and I'm really impressed. Here's some examples for "leather texture, 3d render". In this context "3d render" appears to be a magic incantation that makes the results cleaner. You can give it hints with img2img mode too, just make a textureless brown square in MSpaint and throw that in img2img (with the denoising slider set high) and you'll be guaranteed that the resulting texture fills the whole frame and isn't that weird angle thing.
|
|
#
?
Sep 12, 2022 17:53
|
|
|
"Don't talk to me or my son ever again" "A very tall man-bear hybrid, idyllic, insanely detailed, realistic, ornate, black velvet, hard shadows, diffused lighting, vivid color, post-processing, massive scale, stunning environment, 4k, uhd  Sometimes you just get the perfect hallway. "The_hallway_to_the_land_of_the_dead,_skull_motif,_lava_pouring,_idyllic,_landscape,_dark_lighting,_insanely_detailed,_crystalline, ornate, hard shadows, diffused lighting, vivid color, post-processing, massive scale, stunning environment, 4k, uhd 
|
|
#
?
Sep 13, 2022 04:42
|
|
|
AARD VARKMAN posted:i could easily imagine a year or two from now being able to link a game to a fine-tuned model hosted online and use it to generate all of the graphics. I think for KKKrieger type local game asset generation the whole point would be to use the local GPU. A much smaller model than SD could be trained on imagery specific to the game - textures, landscapes, characters, vehicles, etc. Create textures from that model with specific seeds and prompts, then all you need to provide on top of that is masks, mapping and modifiers. I bet terrain geometry can be generated in a similar manner, if not now then soon. The textures and geometry would be cached locally once generated. It could make generation and distribution of assets for new campaign maps and dungeons extremely fast and cheap in terms of bandwidth. have a cursed img2img experiment
|
|
#
?
Sep 13, 2022 05:01
|
|
|
If you have a 3090 you can grab Textual Inversion, and train it on any dumb poo poo, and it'll be loving awesome.
|
|
#
?
Sep 13, 2022 05:05
|
|
|
  What did I do wrong on the SD upscale to make the poor king grow a new hand? too much denoise? I love a lot about the second image, but the details in the original are so much better
|
|
#
?
Sep 13, 2022 06:06
|
|
|
Been experimenting with surreal compositions, got some interesting stuff  
|
|
#
?
Sep 13, 2022 10:03
|
|
|
is there like a holy grail website or reddit post that shows how best to get these things working on your pc at home? I tried a plug-in for Blender that uses Stable Diffusion but it’s fail
|
|
#
?
Sep 13, 2022 11:24
|
|
|
I can't help but feel that its crazypants that we can improve what we get through all the stuff *surrounding* the SD AI. Its like we barely understand the very thing we created and are still figuring out its capabilities. I mean, all these new features are still running on the same core model aren't they? Do any of the people who create these AI models even know how they work once they're fully formed, or is the end result a black box we just can't pierce and yet still wield? Do we even have the means to properly peer into them and discern what goes on in there, what it knows and doesn't, and how they work?
|
|
#
?
Sep 13, 2022 12:21
|
|
|
Mindblast posted:I can't help but feel that its crazypants that we can improve what we get through all the stuff *surrounding* the SD AI. Its like we barely understand the very thing we created and are still figuring out its capabilities. I mean, all these new features are still running on the same core model aren't they? Generally though I'm pretty sure the people who created this do understand how things work. It's just still a fairly new area and you can't figure out everything immediately.
|
|
#
?
Sep 13, 2022 12:27
|
|
|
echinopsis posted:is there like a holy grail website or reddit post that shows how best to get these things working on your pc at home? This one https://github.com/AUTOMATIC1111/stable-diffusion-webui Seems to be the easiest. Follow the instructions under Installing and running Note that you will also need to install Nvidia's CUDA toolkit, which in turn will whine at you to install visual studio so it has a compiler (midway through the installation of cuda it will say it needs visual studio installed, at that point you can download it and get whichever version it asks for).
|
|
#
?
Sep 13, 2022 12:53
|
|
|
Tunicate posted:This one Thanks mate
|
|
#
?
Sep 13, 2022 13:09
|
|
|
Mindblast posted:I can't help but feel that its crazypants that we can improve what we get through all the stuff *surrounding* the SD AI. Its like we barely understand the very thing we created and are still figuring out its capabilities. I mean, all these new features are still running on the same core model aren't they? There are researchers working on this https://www.youtube.com/watch?v=AgkfIQ4IGaM Analytic Engine fucked around with this message at 19:04 on Jul 28, 2023 |
|
#
?
Sep 13, 2022 13:46
|
|
|
We have invented the gear, we understand they understand the uses in how they designed it, in the hands of others. Weird things have been discovered in how to use it better. Users often find faster ways to use software and have ideas to improve workflow that devs don't though. Devs don't use they software they make the software. Much like the gear art AI seems very simple at first but when combined and interconnected can make extremely complex systems very fast. by simple I mean to use we are seeing this now with the backlash about taking artist jobs. Yes I can put in Santa riding over New York City and get my Christmas card just like that but there's so much more to get one worth selling, one at will pop vs all the others doing the same.
|
|
#
?
Sep 13, 2022 13:59
|
|
|
I asked Open AI to write a law for mequote:A law to criminalize painting poles a different colour to other poles. I asked for a law to criminalize painting poles different colours, but the resulting law seems pretty elaborate and requires two people to conspire to paint two separate poles different colours with intent. I think this is a good law, and I would vote for anyone who campaigned on this, because gently caress it, politics SHOULD be confusing. duck monster fucked around with this message at 14:05 on Sep 13, 2022 |
|
#
?
Sep 13, 2022 14:02
|
|
|
Tunicate posted:This one You absolutely do not need to install Visual Studio to use SD. Cuda will whine at you during installation, but you can ignore it.
|
|
#
?
Sep 13, 2022 14:10
|
|
|
I'm finally trying to install SD for use with my AMD card following this guide https://rentry.org/ayymd-stable-diffustion-v1_4-guide, but I can't seem to access the huggingface CLI from my command line. I just get the error message "The term 'huggingface-cli' is not recognized as the name of a cmdlet, function, script file, or operable program." Can't seem to find anything online about it
|
|
#
?
Sep 13, 2022 14:31
|
|
|
Boba Pearl posted:If you have a 3090 you can grab Textual Inversion, and train it on any dumb poo poo, and it'll be loving awesome. Hi yes it's me, I'm working on a PC-98 one right now. I doubt it will work as well as I hope because the resolution in SD isn't high enough to get the dithering right but gonna let it chew on it for a few days and see what comes out. If its any good I'll see about getting it posted somewhere. edit: Burning Rain posted:I'm finally trying to install SD for use with my AMD card following this guide https://rentry.org/ayymd-stable-diffustion-v1_4-guide, but I can't seem to access the huggingface CLI from my command line. I just get the error message "The term 'huggingface-cli' is not recognized as the name of a cmdlet, function, script file, or operable program." Can't seem to find anything online about it It looks like they expect the huggingfaces-cli to be installed with the transformers package. Maybe check and make sure that package installed successfully, that you started a fresh console afterwards instead of reusing the existing on, and that if you are using a virtual environment you have it activated in that terminal. Also looks like they wanted you to run part or all of this in an Administrative command prompt, that user might have a different Python target. If you check the python version in your normal shell do they match? All I got but might help. Objective Action fucked around with this message at 14:44 on Sep 13, 2022 |
|
#
?
Sep 13, 2022 14:36
|
|
|
Objective Action posted:Hi yes it's me, I'm working on a PC-98 one right now. I doubt it will work as well as I hope because the resolution in SD isn't high enough to get the dithering right but gonna let it chew on it for a few days and see what comes out. Thanks, I'll see what I can figure out, although I feel I'm in over my head here.
|
|
#
?
Sep 13, 2022 15:02
|
|
|
The SD tiling feature is so good. I wonder how it might be possible to train a model to output texture channels like specularity, bump maps, normals, etc.
frumpykvetchbot fucked around with this message at 02:56 on Sep 14, 2022 |
|
#
?
Sep 13, 2022 17:27
|
|
|
frumpykvetchbot posted:The SD tiling feature is so good. I wonder how it might be possible to train a model to output texture channels like specularity, bump maps, normals, etc. Possibly, but that might be overkill. You can already create other maps from the diffuse map with https://boundingboxsoftware.com/materialize/
|
|
#
?
Sep 13, 2022 21:16
|
|
|
Progress update, I asked my card to spit out a matrix of 4096 images with a bunch of prompt sauce and some of my in-training models to see how its going with the PC-98 model. As expected its not really able to do the dithering right due to the resolution, but to my surprise it has already figured out to add shoulder pads and 90s hair to people among other things. Weirdly it insists that people should have purple or white hair despite only having one person with purple hair in the training set and no white hair ones. Will post some examples when this finishes in a few hours and I comb through the output for good ones.
|
|
#
?
Sep 14, 2022 00:13
|
|
|
by the way, did I post this?
|
|
#
?
Sep 14, 2022 01:12
|
|
|
quote:*Celebrity name*, as hulk, digital painting, artstation, concept art, smooth, sharp focus, illustration, art by artgerm and donato giancola and Joseph Christian Leyendecker, Ross Tran, WLOP -H704 -C 10 is a source of endless joy. Caveat warning: The output for female celebrities gets quite horny pretty quick. Complete Gallery with highlights - possibly  - here. I'll probably keep updating the gallery until I get bored.Edit: Got a bit carried away, 2 hours later its 50+ images and counting  You wouldn't like them when they're angry.     Tatiana Maslany lol  And finally, just to make sure I'm going to hell. 
mcbexx fucked around with this message at 12:24 on Sep 14, 2022 |
|
#
?
Sep 14, 2022 01:40
|
|
|
try slapping in a bunch of poo poo like "EXPLOSIONS BOMBS MISSILES TRACER ROUNDS RUINED BUILDINGS"
|
|
#
?
Sep 14, 2022 01:54
|
|
|
instructions unclear, I think I made a mistake edit: by the way, I also made this Automotive Insanity thread. Put car pics there so I'm not the only one posting. https://forums.somethingawful.com/showthread.php?threadid=4011793 Jonny Nox fucked around with this message at 02:25 on Sep 14, 2022 |
|
#
?
Sep 14, 2022 02:15
|
|
|
|
| # ? May 30, 2024 10:10 |
|
|
computer, give me an excuse for being late to work
|
|
#
?
Sep 14, 2022 04:08
|
|
