INSTANTLY improve your AI art with MODELS and LORAS
ฝัง
- เผยแพร่เมื่อ 22 ก.ค. 2024
- Full guide for Stable Diffusion, models, loras, textual inversions, embeddings and more! Fit for complete AI art beginners and AI masters.
Create exactly what you need by teaching Stable Diffusion.
------------ Links used in the VIDEO ---------
Unlimited AI images, models and more!!!: civitai.com/
Ivory.
civitai.com/models/21703/ivory
Realistic version
civitai.com/models/4201/reali...
counterfeit and abysshellmaple
civitai.com/models/4468/count...
civitai.com/models/16727/abys...
Deliverate Version 2
civitai.com/models/4823/delib...
RevAnimated
civitai.com/models/7371/rev-a...
easynegative
civitai.com/models/7808/easyn...
epiNoiseOffset
civitai.com/models/13941/epin...
PixHell lora
civitai.com/models/21270/pixh...
LeagueofLegends
civitai.com/models/22112/leag...
------------ Social Media ---------
-Instagram: / not4talent_ai
-Twitter: / not4talent
You want to unleash stable diffusion even more??
You will need this then:
• Get FULL creative cont...
Make sure to subscribe if you want to learn about AI and grow with the community as we surf the AI wave :3
#aiairt #digitalart #automatic1111#stablediffusion #ai #free #tutorial #model #lora #betterart #goodimages #sd #digitalart #artificialintelligence
0:00 intro
0:20 Models explained
0:50 How to pick a model and download
1:24 Examples on models strengths
1:49 VAE + where to install
2:19 Model install
2:28 Limitations of Stable Diffusion
2:48 Surpassing the limitations
2:53 What are Loras/Embeddings
3:10 Differences between the two
3:20 Install and use Loras
3:42 Read the recommendations
4:05 Make a text file to remember XD
4:19 Embeddings and how to use
4:38 Training quality
5:00 Do they work?
5:15 Delivering on the promised
5:43 My personal favorits
![[LIVE] : ONE ลุมพินี 71 วันนี้!! คู่เอก "ทรงชัยน้อย vs รักษ์"](http://i.ytimg.com/vi/_AGPnq2p6Js/mqdefault.jpg)
Thank you very much. Your tutorials are straight to the point and clear. I've seen other tutorials that are too long and then didn't even get the information that I wanted. Keep up the good job and more please 😁👍👍👍
Thank you so much!! comments like this help a lot! Next up, inpainting and image to image
jinx from fortnite, please be a running joke on this channel LOL
Hahahhaha I tend to do that yeah😂
I was wondering if that was on purpose or misshap
thanks for the explanation and your cat is very cute 🥺
Hahahahha supa cute😂🥰
Great job! Your explanation is very clear and understandable! Keep it up!!😊😊
Thankss! I'll try for sure!
keep creating the details of prompting bro, esp for SD, love it
Lots of tutorials in mind, but will do, thank you!
You are awesome ❤
hahah TY for the super wholesome comment, you are awesome too!!
extra networks button has been removed, and there is tab next to negative promt instead
You're so good at this. By this I mean a I art, Giving the most information in the shortest amount of time without being confusing, keeping it entertaining and In turn earning my attention, your knowledge either conscious or unconscious that surprise is the foundation of humor, and expressing that wisdom, in a way in a way that wasn't overbaked* so it never felt like you were pandering for and laugh, forgive my assumption* your mastery of more than one language *assuming English is not your first language. Last but not least is you're a good teacher. #HighFive
I feel the need to say this just to stop any negative comments but I feel youtube is a place where people can point out everything that's wrong but nobody ever points out every single thing that's right so that's why I was excessive in my standing ovation
@@rhythmoftherain5341 wow, this is probably one of the most supportive comments I ever got. No words.
I can only say thank you so much for the compliments and Im really happy that you found it entretaining and useful. Hope to keep it like that.
The support on the comments has been amazing, really nice community. So thank you all very very much as well.
Hope you have a nice day, you made me start mine in a really nice way! Ty!
Your tutorials are incredibly accurate, helpful, and appreciated. However, I had to downgrade the playback speed to 0.75 to absorb the content volume. My Adderall prescription is almost ready ;)
Hahababahhahahahhahahaha thank you!!!
I agree, started editing my vids like they where shorts for tiktok. Now im dialing this down a bit, as they where fast af.
Still kept the ammount of info x vid too high imo, so now im dialing that back too, deviding the info into more vids.
Adhd gen aint yet looking for SD tutorials 😂
5:10 Did you say jinx from fortnite? D:
😂😂 yeah, intended as a minijoke xD (the prompt is written as jinx, league of legends, tho hahaha)
Is it still very hard(almost impossible) to train a Lora with a 2060 gtx?
Locally I think so, Even though looking online there are people that have some options. I can't personally attest to them though.
Also this video's second part uses a google colab, to train a Lora. I think that is accessible for anyone: th-cam.com/video/7m522D01mh0/w-d-xo.html
It is also a really good tutorial!
hi bro, not all my lora are showing on SD in a1111. I copied them all on models/lora folder from civitAI, any suggest? 2 or 3 are missing :O
If you recently copied them. Try hitting "refresh".
If they are old. Maybe check civit AI to make sure they are not lycoris instead of loras. Or embeddings.
If nothing like this works,maybe the files didnt download correctly
@@Not4Talent_AI I'll try to download them again, but it's very strange because I've always downloaded from Civitai, they are up to date and only 3 out of 11 lora do not appear when I click on "show/hide extra networks" on A1111
Wow the pic of jinx is nice :p
Hahahaha ty! (Or the AI says, I guess 😂)
I really wish you'd give a tutorial on how to actually train LORAS and Textual embeddings
I will! It is a little down the line though, have other tutorials planned, if you need it rn I can give you some cool videos that helped me when doing it
@@Not4Talent_AI please do
your videos are really helpful and easily understandable
@@chigbundu Thanks!! I'll put it on my list then 😄
I've been watching your videos for about a week, and I think I've learned more in these last 7 days than I've learned in the last few months using stable diffusion.
I didn't really find a video on this channel that talks about the correct use of Loras (this one is the only one, but it's only the basics), you know when combining multiple Loras makes non sense images, for example poses with characters will not always work, or why some Loras recommend using HIRES.FIX (here with potato gpu 4vram it's almost imposible!), the difference between using hires.fix and img2img, etc.
A video like this would be excellent imo (im struggling here lol), There are thousands of Loras online, but many of them struggle combining it with other Loras when only knowing basics, Controll Net is the only posible solution that it comes to my mind rn, maybe is there something easier?
Thank you so much, your content is pure gold
Thank you so much for the kind comment and for the idea!! I'll look into it :3 I have to test some stuff myself but I think it is a pretty nice topic, tyty! (have 2-3 vids before I can really get into this one tho. But Noted 100%)
PD: some use highres to restore the faces. some times LORA training creates weird artifacts on the face if not trained properly or without enough resolution on the dataset. So using high res fixes it almost every time
@@Not4Talent_AI Gracias genio! Me quedó mucho más claro ahora, vi que algunos Loras lo recomendaban pero nunca aclaraban por qué, y siempre que podía evitarlo lo evitaba, pero no siempre era posible :(
@@bakugameplays otra strat es generar normal sin highres y ver si peta algo. Si la cara sale pocha luego mejorarla en image2image
:3
@@Not4Talent_AI Siii eso es lo que suelo hacer, solo que algunos Loras a 512x512 por alguna razon salen todos deformados, pero les pones el hires y los arregla.
En fin, creo que es más un problema de los mismos loras, mejor hacer un baking de +40imgs y alguna la va a pegar para el img2img 🤓.
Hacer baking con hires fix para ver si alguno la pega es inviable en todos los aspectos, por eso debe tener mala fama
How about a tutorial about training your own anime LoRA?
Yess, it's a little far up the ladder though. Have other things planned first. If you need a tutorial for it rn,Olivio Sarikas has a pretty nice vid on it!
@@Not4Talent_AI I'm struggling with it. Kohya committed a new update to additional networks and now I can't create LoRA, I tried and I get a black screen
@@andrewjaeger Ohhh, I was unaware of that... Is there a specific error on the cmd?
Btw, this guy makes good videos on training for style, maybe it will help: th-cam.com/video/Icf3ZQ67KPI/w-d-xo.html
Not sure if it will fix the error that it is giving though
@andrewjaeger fresh install. That's what i did.
No big deal.
I had issues with some optimizer too and it fixed my problem.
I bet you found the issue by now but still
For some reason my V2 doesn't work
Hi! I don't really know what you mean by "V2". But if you want I'll gladly try to help with it
@@Not4Talent_AI oh I was referring to version 2 model's like Illuminati, they stop working once I updated my control net. It's fine though cause I can work around it by finding the right checkpoints.
@@chriscodling6573 Ohh I see. If it is fine then cool!! kinnda wierd that ControlNet messes up your SD 2 models though... Don't really know why that would be
that models list at 2:23 is looking a little sus 👀
ahhahahhahahhaha
someone caught it xD
Have no technology knowledge can you make one for me please.
It kind of depends. What do you need?
@@Not4Talent_AI all an ai model needs. To promote brand and be interested in English football, darts boxing. Keeping fit. Blond. Blue eyes smiling. Different posses Different settings. Beautiful. Dancing. Clubs beach wear. Nfw content. Clothing changes.
Poker games
Driving a top make of car.
most of those you dont really need a LORA for. would the lora be for a consistent character that does all of that?@@atruereview7298
How did he hide all his porn related Loras for the video capture? Asking for a friend.... ;)
moving them to another folder xd