Stable Diffusion BEST Tutorial for Prompts, Beautiful Results | Master Prompts for Stylized Art
Вставка
- Опубліковано 28 чер 2024
- Stable Diffusion BEST Tutorial for Prompts, Beautiful Results | Master Prompts for Stylized Art. Master Stable Diffusion get the best results with powerful Prompts. 45 minute in depth Tutorial on how to write the best Prompts for Stable Diffusion using custom models.
Stable Diffusion is a deep learning, text-to-image model released in 2022. It is primarily used to generate detailed images conditioned on text descriptions, though it can also be applied to other tasks such as inpainting, outpainting, and generating image-to-image translations guided by a text prompt.
#aiart #ai #stablediffusion - Спорт
Thank you. I've heard the same content over and over, but the way you go over it makes it a lot easier to understand what the prompts do.
And now I'll always remember the prompt order.
Style, subject, subject details, background, background details, stylizers, artist
Fantastic tutorial Mike! Really looking forward to your next videos.
Hi Mike, I've been following your artwork for a while. Just wanna let you know while you're blowing these images up, it also blows my mind. 😆Thank you for the great tutorial!
Amazing tutorial! I like how well-structured it was and you covered a lot!
What a nice, chill tutorial video :)
Would love to see more of your tutorials about stable diffusion! You take it slow and steady.
yes please, I would love to learn on how to speak to it.
please do some tutorials,
step by step , pleeeeeeease
Excellent video! As someone new to AI art, the world of prompts seems to be shrouded in mystery to me. This is definitely the best video I’ve watched on the subject so far. Thanks!
A cowboy shot is slang, the term is 3/4 shot. Finally film school has paid off.
This video is really helpful! Thank you so much for your valuable tips, I'm eagerly looking forward to the next videos.
This video is fantastic - this is what i was looking for - thank you
Looking forward to your next video
Thanks for the extremely helpful and well produced video!!
thank you for this excellent tutorial - can't wait for your next video
Looking forward to your future videos!
Absolutely fantastic tutorial, thank you very much! my AI generation has become much, much better now 😊
Best tute I've seen on this topic. I hope you do more.
Glad I found this video. I think it's very underrated. I would love to have more of this. I'm not that keen on the livestreams, they are too lengthy. I would rather see shorter (scripted) tutorials that go more in-depth into one specific topic.
This should be required viewing for every new SD user. Spread this video far and wide!
i love how everything else on this channel is yoga ^^
good luck with the ai tutorials, was a good presentation
I sat down and watched this whole video so good.
Awesome tut ... it helps me a lot to understand Stable Diffusion use of keywords and style structures. The "beauty pinup" model, is that one of your making or is there a place to download it?
Hi, great tutorial! Where can I download the model/checkpoint you use? Thanks!
Great video, learned alot :) where did you het the Beauty Pinup checkpoint by the way? just a lovley one with great results :)
Really Great video TY. This is the third I watched on this subect and I kind of felt like Goldilocks. The first was to little, the second was to much but this video was jsut right. Thanks again, my images have never looked better.
This is weird. I thought this is AI lesson channel but turns out to be yoga lesson channel... lol
That was 4 years ago lol
Though the same XD
@@150_asharifadhilaha2 i think you should make a new channel 😅
Yeah, happens a lot lately.
My favourite comedy channel recently changed name and became a fishing channel.
One of my own channels recently changed name and is now a on-and-off gaming channel.
Even though I'm one of the ones who did it, I really couldn't tell you why so many does it right now. Guess we all just wanted some change at the same time...
amazing, keep it going dude.
Great info and tips, but you forgot to mention what checkpoint you were using!
You can see it in 7:27 , it seems to be "Beauty Pinup" but idk if it is made by him
Hey Mike, this is Komar, this is really cool stuff! Very impressive, I'm thinking I may just want to dabble with it myself, it really is pretty neat. As for your video here, quite informative, good job!
I'm following along and notice your computer is way faster than mine. I thought mine was powerful. You clearly have something better.
One thing i learned from AI from using it for just a few days, that i will start using in real life, is to be specific not only in what you want from something, but to be really mindful of what you dont want.
Actually, without knowing, was using this way of looking at things that i stoped being atheist and believing in something, only after clarifing what wasnt god that i foud what god really is.
"Unwanted" prompts is key to have good result.
Would you be able to share more about the software you use? I'm testing Easy Diffusion v2.5.39 right now and it takes twenty minutes to generate five images. Are you editing out wait time for the video or is there user error and I'm missing something?
Crap I was hoping you had more videos for this but all I see is yoga lol. This video was very helpful to me, thanks
Hi! Bro what is the best Sampling Method to use?
Really good structured introduction.
Great detail, process and workflow. It's on everyone's tongue -- you are signed into a server on stable diffusion -- and can we go there to access that UX? Thanks for spending the time for us.
He's using localhost which means he's running it on his own computer. The site rundiffusion has machines you can rent by the hour for $2.50/hr and those have the same UX you're wanting, which is just the Stable Diffusion UX really
'Getting a lot of head shots, didn't ask for that...'
Yes, you did. That's what portrait means.
"a painting, drawing, photograph, or engraving of a person, especially one depicting only the face or head and shoulders."
Great video, I work with 3d rendering because I can create scenes in images for example two people playing tennis. What I'm wondering/waiting for is can AI create an image like that, basically characters interacting with each other?
Thanks for the video and book reference. I am better educated as to how this or any artwork is created with this computer program.
She like the lace Sir, give her one or something! xd
Anyway, nice tutorials help a lot. thank you guys, and have a nice day. :D
43:34 I actually remember and come back to this point
so is there a furture video that would go over how to correct full body shot faces, yet?
Because the image generates so quickly, what are your CPU specifications? Please let me know with the available link if there is a video of yours where you explain how to make an image thus quickly.
How do we use the model that you used? Thanks.
How do I get the full version? I can't seem to find it.
I used exactly the same negative prompts, and promps, with SD1.5 and the results are horrible, which model are you using? couldnt find it out there
same issue
wow, can do another one? wuld have listen to 3 hous of this. :D
what model are you using, i can't get the same quality results
Thank you for the vid, but there is a question - what settings are you using in your UI? I tried all the same prompts on Playground and DreamStudio with the very same settings (sampler, scale, steps, negative prompt) - but my results were never even close to what you got (almost all of your girls had pretty faces but I had random faces and not so beautiful)
I noticed the same thing, then noticed he is using a different checkpoint so that is probably why
M going to try "inside a bottle" not "in a bottle" see if any change, many thanks for tips, I'd kinda worked it out but the underscores and syntax should help many thanks
how do I access that stable diffusion tab you are using?
He's running it on his own machine
what model are u using?
which gpu are u using ?
The model youre using looks like the openjourney model from prompthero. Am i right?
how do I access all your stable diffusion videos
Now can you tell us how to install chkpt you used.
hello, is it possible to make the AI-Generated Images (ULTRA REALISTIC) move and speak like a real person through Prompting?
Yes and no.
As far as im aware you can not use stable diffusion for that. BUT
There are other AI programs for that
what site it use?
Mike, I've been following your work. Your pin up girls are all very impressive (even if a tad on the over-busty side). I know that you have created your own character models for those pinups. It seems to me that until near the end of the video that your faces all seemed to be the same woman (or nearly so) that you do in so many of your pinup images. (That pointy chin is a real giveaway.) Did you have that custom model influencing what we see here? I ask because it seems that, not only are the faces nearly all the same, but there seems to be no ethnic diversity in the prompt "beautiful woman" other than white females. I am not knocking the results. I am simply trying to determine if what we witness here is a result of the basic Stable Diffusion modeling, or rather, your own set up. (That is to say, if I want an image of a beautiful asian woman, do I need to add that specific modifier? In your experience, is Caucasian the default definition of "beautiful" in Stable Diffusion's training? Or is it a reflection of your own modeling?)
Also, I certainly hope that you cover negative prompting soon, because I think anyone trying to imitate your results without adding the negative prompts that you have listed will be very disappointed by all of the inexplicable ugliness and distortion that is so common when using Stable Diffusion.
And-while I have your ear-do you think that your workflow will work with any of the Stable Diffusion-based generators out there? I use Playground AI. Or are these techniques more limited to running Stable Diffusion via a local install on your PC?
BTW it was great to get your insights on how all this works. Please keep these tutorials coming!
The negative promp is: nude, naked, nsfw, lowres, text, error, cropped, worst quality, low quality, jpeg artifacts, ugly, duplicate, morbid, mutilated, out of frame, extra fingers, mutated hands, poorly drawn hands, poorly drawn face, mutation, deformed, blurry, dehydrated, bad anatomy, bad proportions, extra limbs, cloned face, disfigured, gross proportions, malformed limbs, missing arms, missing legs, extra arms, extra legs, fused fingers, too many fingers, long neck
And about the "beautiful woman" you mentioned is really interesting.
"This is a known "bias" in the current models, not just Stable Diffusion, but DALL-E and Midjourney also have similar biases. Other biases exist too, like asking for a CEO would most likely return a white man, and asking for a nurse is almost always a woman. It's not intentional; the models took images from public sources, and the biases are in those sources.
There are various efforts to update future models to more accurately reflect the variety of people."
So, I made a test using the prompt "beautiful women" and using the same negative prompt as the one in the video. After generating around 50 women It appears that around 15 to 20 percent of women generated are non caucasian women. But almost none are asian or really have dark ivory skin color.
may be a dumb question. but how do I even get stable diffusion? no links seem to work when I search it
My dude that honestly is not an easy question to answer in youtube comments. Message me if your still curious
Great explanation
i tried loads of prompts to try to get a dog to run on a rainbow, but the rainbow is always behind it. what do i need to change? my prompt is "black and white pitbull, running over rainbow_bridge"
Am describing one character, but it keeps giving me multiple characters in the same frame, even when I specify I only want one. (Any help, would appreciate)
how fast is your PC tho 😆😍😆😍
How can we install that in our machines?
Use the install guide? lol
Wait.
Are we Yoga or AI Art Prompt Engineers?
I used numerous AI tools. Blue Willow is the one I like the most. Using Blue Willow, I produced stunning photographs, and it's really cool and most importantly it's completely free
I made the speed 2x to keep focused
Please give me the url for this interface to follow your tutorial
I don't know if YT is seriously bugging out right now, but when I click on your channel, and your videos, I get this video then LOADS of Yoga Pose videos made 4 years ago....
Color bleed is a real problem. maybe there was a problem with hardware at the beginning, but really we should have advanced to the point that 'red car' should be identifiable as a single concept and not treated as 'red' and 'car' separately applied. Making the AI "smarter" but still too dumb to keep adjectives attached is not great. What's weird is it can generally get compound concepts in other contexts like "night sky" or "old tree".
gigo
You jump right into it without telling which model you are using... could you please let us know? Thanks!
1.25x perfect playback speed
most important lesson: Use nouns instead of adjectives, because adjectives bleed and nouns don't.
How, how, how???? I only get bad results.
Also, this channel has only ONE AI video and full of yoga videos with different girls, but people call him "Mike" and saying (a lot) that they are following his work?
Some shady sus shit.
Why is stable diffusion checkpoint line missing from video? A little suspicous to me. You're obviously hiding it.
His "cool" is totally a "yeah, sure, who gives a shit, let's move on with what I consider important stuff"
You didn't say which model you use
Guy teaching yoga suddenly decide how to do AI art😂😂😂😂
Curious why it doesn’t include any images of Black women?
You keep describing the function of the AI as if it is finding images containing the things you describe in the prompt, but that's not how it works at all.
You even implied that it "contains" pictures when you said, "the AI is not gonna have a lot of pictures of women in bottles in it."
The details regarding prompt formatting and structure are good, but if you're using this kind of language to keep it simple, you should probably stop, because it only reinforces the myriad of misconceptions regarding how these AI models work.
Dude the woman at the background is keeping creeping me out
I type totally the same thing, in prompt and negative prompt, and i do not even get something that looks like a women :/
You realize all these beautiful women are white? This is an issue with AI, it often copies human failings.
Omg let me guess. The creators or the input of info (feeding) into the creation of all these ai programs are men...everywhere I look on UA-cam it's predominately women, women's sexual parts.....being generated....WHO WOULD HAVE GUESSED! Good video info with prompt creations....thanks....
That something that hasn't changed in the last 10'000 years 😂
just want to point out. you can generate ai images all you want. you will never get a job out of it. because ai generated images can not be copyrighted. meaning No real business is gonna pay you too generate or prompt images. as for a buisness too thrive they need full exclusive control over their concepts characters and designs. and the only parts that can be copy righted is any significantly modified parts of an ai piece. in other words. ai art is a dead end. you are not an artist if you rely on a machine. and prompts are not creative. it is equivalent to giving directions to an artist you commissioned.
Wow