Turning an Illustration into a Full Monster Character with Stable Diffusion AI (Walkthrough)
Vložit
- čas přidán 29. 09. 2022
- You asked, I delivered: no timelapse - this vid is realtime except when SD generates pics. Watch me do a full walkthrough of turning a cute monster drawing into a kid's 3D movie style artwork!
Thank you Steffi Engel for illustrating the original image! Visit her store with the code SOFTUNICORN for 15% off: www.pummeleinhorn.de. Works in EU + England, Norwegen, Switzerland, and Liechtenstein. Minimum purchase is 5€, expires 31.12.2022.
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...
----------------------------------------------
Did you like this vid? Like & Subscribe to this Channel!
Follow me on Twitter: / albertbozesan
I genuinely think that you are "actually" the first AI artist on YT, that spread information in a direct and ethical way, great content as usale!
That’s a big compliment, thank you!
You're really showing AI as what it is: not something that creates art on its own, but a powerful tool.
Love that.
You can put parenthesis ( ) around words you want SD to focus on, and brackets [ ] around words of less importance if they are dominating results. The number of parenthesis and brackets also denote strength, (( )) [[ ]], (((( )))) etc. Also don't forget the lower input field, that is a negative prompt, I find that putting meme in the negative prompt gets better results, the dataset that SD uses has a lot of memes in it. text is also a good negative prompt to use most of the time. edit: and of course I continue watching and see you use this! :D
Good tips, thanks!
You're really carrying the AI art tutorials right now, keep up the great work!
Thanks so much! 😄
Wow! ❤ Amazing work, Albert! Thanks for the collab! Can‘t wait to see and hear more of the lollitroll soon. 🍭😏
dude thank you so much for opening my eyes and telling me proper way to utilize the power of a.i
Thank you for your wonderful and useful content, Albert.
Thank you very much! I like your videos a lot. Please keep it going!
Will do! Thanks 😄
@@albertbozesan vielen Dank, ich wusste nicht, dass du ebenfalls Deutscher bist. :)
Great work as usual and a really nice explanation of the process!
It's really pretty!
Again a nice video, thx
great tutorial thanks !!
great video, i've learned a lot, thanks
I would appreciate the before and after shot side by side at the end of each video like this :)
The way diffusion models work is by learning to remove artificial noise and leave behind a coherent image. So I suspect adding noise to your img2img input would confuse the hell out of the AI 😂
That's so impressive and your explanations are great. Glad that I've found you.
Could you maybe show, if it's possible to show the character then also from different perspectives (e.g. from the side) what would help to make a 3D model of it.
Thank you, glad you like it!
Other perspectives wouldn’t be super easy to do, but I will sit down and try to figure out how :)
@@albertbozesan cool, looking forward to that 🙂 Maybe one thing what i guess could help would to have them in T-Pose (or another "easy" to reproduce one)
Pokémon you're thinking of is Pollywag!
The pokemon you were referencing @33:12 is Poliwhirl I think
It was! Thanks :)
Yap, the photoshop plugin include local processing now. As far as I understand, you need to have adobe creative cloud with the latest version of photoshop. If you can install this plugin bypassing creative cloud, on regular people photoshop, please share how you did it.
Love.
By the way, you are sawing relevant content, thank you for your work.
Nice tutorial brother sorry I need to ask you if soft can be installe on Android?
Not locally because it requires a strong GPU. But you can use one of the online services like Dreamstudio.
Hello Albert, is it possible to make this Monster to a 3D model too? This would be fantastic!
I mean, yeah! It’s not easy, and there aren’t great AI tools yet, but with lots of manual work of course 😄
@@albertbozesan I can do the manual work when you do the first Converting Act to 3D Model! Your fantasy Monster looks very great!
What your dreaming of is Dream Fusion.
Can I use this if I have AMD ? I love that setup easy to use looks like
If you're ok with a few extra steps you could try this guide -
/watch?v=Lk2syPsVMQM
The problem is that Stable Diffusion uses a specific type of processing that Nvidia cards are "built" with. AMD cards are "built" differently, hence the need for a work around.
Immediate like for respecting Greg Rutkowski. Not only does he not want his name used, people are using his name lazily. Do they even know who he is?
Hmmmm, he's got to have a name. How about Vennlig?
Ha! Love it 😄 so it shall be
I've tried running Stable diffusion on my computer without a graphics card...
Got pretty far with the Vega 11 integrated GPU.
But HIP doesn't work out of the box with integrated.
I'm sure it Could possibly work if you were to compile older OpenCl ROCm for your system. (pre 4.5 version)
But yeah.
ROCm support is quite lacking for the integrated graphics.
I'm sure it would work on a RX 480 and up.
But definitely something I don't recommend trying.
Hoh well.
Guess if I want to play with these new toys, I'll be needing at least RDNA 2.
There is a CPU version
I do not see option "Redraw whole image" in my UI. Has that changed recently in update?
Are you maybe in the inpainting tab?
@@albertbozesan no I checked all tabs. I do not see that. Suppose if you ever knew the reason, please reply
@@rajeshkannanmj what happens when you use it? Is it possible that's it's just on by default?
@@rajeshkannanmj Oh, and are you sure you are using the Automatic1111 UI? There are several installations that look quite similar.
@@albertbozesan yes I am using that automatic only
how to instal?
Check out my other video for an easy way:
czcams.com/video/034pKa0Kuu0/video.html
🥰🥰🥰🥰🥰🥰🥰🥰🥰🥰
can i use Albert Bozesan in my prompts as an artist inspirtaion? xd
Haha I wish!
@@albertbozesan Try it. The other day someone said the name actually doesn't even matter...
@@owlmaster1528 true. Just a “by” or “in the style of” in the prompt will help the quality because SD will search its model for reference “by” someone - making it more likely to pick artistic work
Poliwag
YES! Thanks.
My gosh, if I was a good enough artist that an entire community, (especially one poised to take over multiple industries) was slinging my name around, and emulating my style, I would be all about people using my name! I mean, you couldn't BEG for better publicity...
I think a lot of people who are against such things don't really understand how these AIs work, so they see it as the AI copying/stealing their work, instead of the more accurate interpretation that the AI is inspired by their work.
Either that, or they just think "if anyone can emulate my style, then no one will pay me commissions" 🤷♂
@@IceMetalPunk Exactly! If art is your passion, no amount of automation can take that away from you. If art is your job, use it to speed up and improve your work flow.i think folks will see it that way soon though, one they wrap their head around it.
The thing is, I could hire a bunch of people on fiver or whatever to study all of somebody's art, then give them a sample image and say "make this image in that person's style," and get back a handful of results from those people I hired.
That's *all* Stable Diffusion is doing, but it's using a GPU to make the comparisons instead of using people's brains for hire.
If your art is *ever* seen, your ideas have been "stolen" just as effectively as if it was used in the SD data training set. Zero difference *whatsoever.*
@@TherronKeen Exactly. And the inverse is true too. Every artist is inspired by the world that came before them. Maybe not to the point of "Draw two rabbit dueling each other in giants robots in the style of Thomas Cole" like we could get from your example, or stable diffusion, but inspiration and emulation is innately human.
how to make humans?
greg rutkowski is a CHUMP bro… 🤡😂😂
🤷♂️ if he’s willing to speak to press to ask for us not to use his name, I’m not gonna. I think it’s his loss though, I had never heard of him before and really appreciate his OG art now.
And his claim that he can’t search for his name anymore isn’t true. Google shows his art front and center, AI results are way lower down.