How to Make Pixel Perfect Art with AI (Free and Easy) - Stable Diffusion Tutorial 2022
Vložit
- čas přidán 18. 06. 2024
- How to Install and Use Stable Diffusion (June 2023) - Basic Tutorial
• How to Install and Use...
Other options for using Stable Diffusion: / dreamers_guide_to_gett...
CHAPTERS:
0:00 - Intro
3:14 - Buildings
9:42 - People
14:03 - Textures
18:44 - Items
Prompts I used for this pixel art:
Buildings: pixel art of a XXX, colorful, warm and cozy, nintendo, pokemon, stardew valley, animal crossing, isometric perspective
People: pixelated full body XXX looking right, character icon, concept art, harvest moon, artstation, 2d game art, pixel perfect, detailed shading, by Paul Kelpe
Textures: pixelated XXX, concept art, harvest moon, pokemon, artstation, 2d game art, pixel perfect, detailed
Items: pixelated XXX icon, concept art, game item, harvest moon, pokemon, artstation, 2d game art, pixel perfect, detailed shading, inventory
----------------------------------------------
Did you like this vid? Like & Subscribe to this Channel!
Follow me on Twitter: / albertbozesan
This is amazing! I've tried generating some game art for my game after watching that other video you made, and it did a fantastic job! However still not good enough for production. It is more than great for placeholders that's for sure! Still playing around with it.
Thank you for introducing this tool to us!
Sure thing! Glad I could help :)
Also, you should reduce the color count to 16 colors or so by switching to indexed color mode before or after resizing. This will really help give it a pixel art look, since most pixel artists use limited color palettes.
Great tip!
Good app to use for this is photo to pixels
Ok thats good, but can this ai make sprite sheet for my character?
This content is top notch and you're ahead of the game on youtube, these tutorials show exactly the real industry use for the AI generated content
Thanks so much!
Definitely not!
This is amazing! Thanks for the advice, I also looked for it everywhere.
Glad it helped :)
The introduction was very helpful, thanks!
Well done again Albert. Keep them coming.
Thanks 😄
I obsessed your videos and I am adding all this content to my pipeline. I am saying this as an art director, Jr artist must work much harder than I was Jr artist. We have to much creative :) But ın same time we have gain tons of time :) Thank you.
That’s so cool to hear, thank you!
I want to thank you for sharing your knowldge, i didnt knew everybody could use stable diffusion until last monday when i watched your video about It. Its simply fascinating, I've been doing photobashing since ever to represent my TTRPG characters and scenes, this technique fits perfectly to it.
Yesterday a guy paid me to use one of my arts , generated with stablediffusion, in his RPG book, isnt this crazy? Thanks again, generating things with AI has been a relief for my anxiety.
That’s awesome! Photobashing works so well with SD. Congrats on your paid art 😄
Excellent timing with the video as I'm about to start on some concept pixel art for a wild west game I'm thinking of creating, one tip you might want to paste your SD prompts into the video description to help people get started.
Good point, will do!
When resizing in photoshop, you can type in the specific size followed by a unit instead of the percent sign.
Thank you for making this tutorial. It is very helpful!
You’re very welcome! Glad it was helpful 😄
Pretty nice, thanks!
Hello Albert, I absolutely love your videos! I would love to see a video of you looking into generating ui / gui.
I have been trying to figure out how to generate ui, just doesn’t do well in that field.
Ty! And please keep going!
Thanks! I’ve had some success making frames for inventory/menus with img2img. I’ll put it in my plan for future vids :)
Best SD video series around, thank!
Thank you!
What a great video. Super well explained, thanks! Any video in your channel on how to record yourself with soft soft? I'm under the
This is good! Just for those curious: I tried the other method before seeing this video. I mad a concept of what I wanted and it just duplicated it instead of making new stuff. So you do need to have it generate stuff to begin with, unless you are just trying to get your own work to be more varied
Hello. Would you like to do a guid about settings in PC version of stable diffusion? For example on 2:55 we can see "settings" tab at the top of the video. There are so many settings.
Not sure why but these videos are super relaxing.
Haha thanks. I’m planning to make some longer vids of me just making some artwork and explaining my thought process, without a tut. Those could be even more relaxing 😅
If u have a doodle in the center or drawing in the center of a full white bg and run it though img2img with pixel art prompt it does really well for just doing center objects and keeping it pixel goodness and the bg stays white to make it easier
Sry engrish not easy
I've seen quite a few of your videos and find them rather appealing as few demonstrate the progression of their images. I'm unsure if you have any experience training a model, but it would be really cool if you could train a model in a bunch of pixel art to get results that aren't 70% stardew valley.
Wait I’ll try that
Instant fan and subscriber here, these videos are so helpful and wonderfully made! Thank you 🙏 My favorite so far is the concept art tutorial. I would be interested to see how you would approach logo generation (could be fun to do for a rock band or something creative).
I’m finding most useful your tips on prompt words. Do you discover these through community chats or do you know a good resource they are gathered? If one doesn’t exist, would be great content to offer. I’d happily be a Patreon.
Thanks so much! I put logo generation on my list for future vids 😄
I find my prompts by looking at posts on Reddit and playing around with txt2img.
Very glad to hear you’d support a Patreon! What kind of content could you imagine there?
@@albertbozesan Awesome! Can't wait to see it. Content-wise, submissions from Patreon subscribers might do really well as a tier. In general, you could do more of this same thing, potentially less edited (so its less work for you and your hardcore fans can pick up more detail). For me it's helpful to watch the same process a few times with different themes to soak in the creative train-of-thought on tweaking the settings and entering prompts. Also, your personal favorite prompts of the week or month could be an easy low tier. I think reddit is wonderful but find myself in the bucket of folks with low patience for it! I also skimmed through these other comments and found a recommendation you gave for a website with prompts, its not yet very user friendly, I think there's space for a channel like yours to do really well with curated "favorites." If people like it, there's an awesome opportunity for you to partner with these prompt-provider sites for sponsorship as well.
That's a big comment for CZcams, but I hope there's a helpful nugget or two! Again, I really appreciate what you are doing and hope to stay in contact.
Warm cheers,
Ali
@@aligator.martian Thank you for the great ideas! Much appreciated. I will try out some things, maybe wait a little for more subscribers here, but then it sounds like something I'll do :)
im from korea, your workflow is amazing..
Stellar tutorial
These videos are awesome!. Using the technique to make custom art for the item info cards I give my DnD players... Aka my kids
Fantastic 😄 that’s a great application of this
Fyi, Stable Diffusion can be fine tuned if you have 20 GB; that means you retrain the last N layers of network (instead of training from scratch). So, you could create a checkpoint that specialized more on pixel art.
Go buy a card
I love all your videos, thx for doing this, this channel should have 100k subs minium
I guess it's not going to take long 😀
Thanks so much! 🙏
hey, really like these videos! The way you use the AI to generate art is art in itself. Your use of the tools reminds me of how artists discovered the best tools to use for creating art back then. It's a totally new format. Thanks for sharing, liking commenting and sharing to boost your algorithm stuff. Take care!
Thank you so much!!
Many texture artists I know have used photoshop's patterns and content aware fill in a bit similar manner for a decade or how long has it been.
@@devnull_ I mean, yeah, that’s how seamless texture are made 🤷♂️ but the source image used to be photo-based. Now it’s generated.
@@albertbozesan Not exactly talking about that, try taking some grunge or organic texture you hand painted (could be photo too...) - then select small patch from it, like a crack running across surface and do content aware fill. It was a good way to break up textures, add new variants and generate chaos. It doesn't work every time, but is very helpful. Of course it can be used for tiling like you said, by using offset then fixing the cross area with seam then offsetting back the same amount.
@@devnull_ sorry, i thought this comment was under a different video! Yeah, I agree that's a great technique.
Hey Albert, i just saw this video and its so awesome to see your process. I am wondering if your german/from germany because it seems your windows language is set to german. Cheers!
Yes, I’m based in Germany :) originally American-raised though!
How did you know I wanted this tutorial!?!?
Hi, I'm just leaving a positive comment for youtube, so more people can see your tutorials. You are doing great and I'm getting really cool results using some of your tips. There should really be some sort of index of keywords, which work well with the Stable Diffusion. Keep it up!
Thanks so much! I think there are several sites collecting good prompts, like prompthero.com/
Did you try out an img2img based on a preexisting character sheet with the animations ? My guess is probably and it just garbles the image.
Yeah exactly. That doesn’t work super well for two reasons: you’d need a high res sheet so details are still visible to the ai, and then the results are all super different. It doesn’t generate reliably similar looks for different angles…
using soft, can't wait to get my hands on it.
Can you use this in commercial games?
What Nvidia do you have? Thank you for the tutorial!
I’m on an RTX 2070S.
@@albertbozesan cool! Thanks 😊
I think you may be a little confused about the term isometric. If im not mistaken its a camera setting for the "angle of flair" so perspective would be most games where your view is at an angle, think FOV. where as the terms ISO - isolate and metric - of measurement. Is the opposite of perspective, where the camera has 0 angle. This is demonstrable by the amount of tiles you see always being the same width as the camera lens and never deviating.
Thanks! I do know what isometric is, I’ve done lots of 3D work in that style. I just found it’s a helpful thing to put in the prompt to narrow down what part of the latent space the AI finds its results, even if it doesn’t give me real isometric results.
@@albertbozesan I guess the distinction I'm making is that squares are also isometric. So everything in this video is isometric.
@@sasori720 I was curious so I did a little research into this. According to wikipedia, isometric video game graphics are not actually "truly isometric". So it just means something slightly different when referring to games :)
en.wikipedia.org/wiki/Isometric_video_game_graphics
This is great, especially for indie developers just starting out! Is there anything like this for music generation?
Not that I know of but I’m sure something will pop up very soon!
Sort of, but its not very good. Give it a few years, and music gen will be way better
Is it possible to use AI to convert character sprites to another pixel art style, like converting capcom style to King of Fighter style?
I would try out ControlNet for that! Take a look at my other tutorials.
nice )
A good phone app to resize and pixelate with color palettes is photo to pixels
You're good.
What are cool ways we can use A.I. for?
Can we make twitch/CZcams assets like
Icons
Intro
Outro
Overlays
Vectors
Logos
Etc.
I want to start using A.I. to create stuff so I'd love a guide
You’ve come to the right place! Leave a sub and you’ll get all the info you need 😄
Do you know Pixel Art Diffusion? It would be interesting if you try it and give us your impressions.
Your content is great and easy to understand!
Nice video - not there yet for pixel art, ok maybe if you have run out of ideas for style/mood, but you are creating more cleanup work for you than needed. Maybe ok for low quality pixel art, but I bet the AI can be trained to create different dithering patterns and single pixel lines, and learn the use of limited palette, then it could be used to that part only too.
I'm currently training my own model on pixel art, but its limited, since it doesnt use text to image. each seperate item needs to be trained. So if you want weapons, you train it on weapons, and it will output pixel art weapons. You do need like 10k - 30k different images of pixel art weapons (the same size) and it takes days to train it
Yeah, it’s not quite there yet
I can't w8 to see what ur channel ll looks like when the whole AI thing gets better.
Me too!
maybe instead of isometric, orthographic might be more appropriate?
when i pass or load a picture into img2img the resizing cropping tool has very strange behaviors. anyone elese experience this problem?
Yes unfortunately it is super broken right now :(
@@albertbozesan oh i understand, thank you very much for letting me know!!!
is it still being worked on or should i expect this to be it?
i will have a look at the issues in the github. i am a programmer myself maybe i can make myself of use at some point.
many many thanks for your cool webui, i have already generated a ton of images without img2img :)
@@arisrayden I am not involved in the development, so I don’t know - but if you want to get involved, I’m sure many people (myself included) would be very grateful!!
Where do you get this Front end for it?
Check the link in the description, there’s a lot of options :)
how i can download stable diffusion,all tutorial i see doesnt work,my pc just dont download it correctly
If you have an NVIDIA GPU on a Windows PC, try this: civitai.com/models/27574/super-easy-ai-installer-tool
the author answer please.
what kind of model is this? and how did you find her?
This is Stable Diffusion with the 1.4 weights.
I don't know, maybe it's a faster way to get a soft than using the soft roll
Any tips for Vector style art?
Yes, check out my other video on making vector art :)
Well once I get my 2080 ti I'll start training an embedding to make SD better for pixel art
can u make a video for issometric assets?
I’d say it’s a little trickier because the angles need to be perfect. Will look into it! And maybe when a better trained model comes out :)
i believe in year or so there will be ai generator for 2d sprite making that would automatically make the idle,jump,directions strip
send you the link of it
how can I generate animation character
Check out this model: huggingface.co/Onodofthenorth/SD_PixelArt_SpriteSheet_Generator
So... It appears you've happened to train the model on stardew valley, some other game and some more stardew Valley
it looks like shit aswell
Would like to see someone train with a Mugen char library.
That first one looks awful identical to Stardew Valley. Ai is just putting stuff together from existing material isn't it? What if the generated piece really is too similar to an existing piece? Could this end in a copyright issue?
It’s more complex than that. The model is trained on millions upon millions of images that couldn’t possibly be stored in the resulting 4GB model. There is no chance it could reproduce something that already exists in the dataset.
@@aliasmcgames it copies watermarks onto images that have never existed, too. It just learned that watermarks go on things sometimes.
I would like to see one for making videos with SD.
What kind of videos? I’m working on a tut now for VFX Makeup using EbSynth.
@@albertbozesan Yes, that would be sweet. I have not seen many for that.
I love your tutorials! And I found...well, not quite a fix for the offensive word in the installation guide. I downloaded the "Word Replacer II" chrome extension for my browser (I'm using Vivaldi), and I added a substitution rule that replaces that word with "reward" on any page I visit. I also considered "record." I would love a funnier suggestion...🙃
Ha! Love it.
yo yo toby fox and creo brought here doe
35 seconds into the tutorial, and I see an AI created Adolf - The Farmer RPG...
Oh no 😅
If it weren't for you, I don't know what to do, I don't want to spend money on it
Just so you know there is an SD training jut to make pixel art.
Great! Looking forward to it 😄
!
to play with tNice tutorials DAW, thanks a million!!!!!
Make a rap song with A.I. Lol
Haha I couldn’t tell if the result was good or not, my rap skills couldn’t be more vanilla.
This is stable diffusion's error and fault for being unable to actually generate pixel art, and it will get fixed in the future. You should be able to use the prompt "16x32 pixel art in 16 colors of a heroic warrior" and get perfect pixel art output. Also there will soon be a model with extra training specifically for pixel art called pixelart-diffusion similar to waifu-diffusion and robo-diffusion.
lmao sa.. xD
KnewvlIh4AaABAg Recording is !!!!
This is gross man, seriously doing stuff like this will be getting people out of jobs if this becomes more advanced
People said the same about photography, and yet there are more working painters today than at any point in history. Sure, some people will lose their jobs. Yet many others, or possibly the same people, will gain new ones. It’s always happening.
Not pixel perfect this video is clickbait
What do you mean? I use the definition of pixel perfect in pixel art. It will be the correct size in any project - given that your engine settings are correct.
Man of honor, he keeps his promise. Ty you for the vid ser🫶
man I missed this kind of tutorials lol. Great work here, thanks!!!