Stable Diffusion - Juggernaut XL V8 vs V9 - Is one better?

Подписаться 12 тыс.

Просмотров 6 тыс.

50% 1

In this video I show how Juggernaut XL version 8 and 9 compare. For information on getting and installing checkpoints in Fooocus check out my other videos. Master Generative AI for Creative Image Generation! 🎨 #Fooocus #StableDiffusion #DigitalArt #ImageManipulation #CreativeTools
Fooocus is a great interface for using Stable Diffusion and is very easy to get started. It also will run on systems with only 4GB VRAM.
I am running on an i5 32GB RAM Nvidia 3070 8GB VRAM
If you find the video useful and would like to tip, RU-vid does have the Super Thanks option. Those are greatly appreciated so I can have my coffee!!
☕ BUY ME A COFFEE: ko-fi.com/kleebztech
JuggernautXLV9: civitai.com/models/133005?mod...
Download Fooocus: github.com/lllyasviel/Fooocus
How to install Fooocus video: • Stable Diffusion - Dow...
Fooocus Beginners Guide: • Super Easy Stable Diff...
Fooocus Upscales and Variations: • Stable Diffusion - Foo...
Checkpoints and LoRAs the Basics: • Stable Diffusion - Che...
Fooocus Poses and FaceSwap - Image Prompts: • Stable Diffusion - Pos...
Zoom with Outpainting: • Stable Diffusion - Zoo...
Inpainting with Fooocus - Don't Regenerate, Fix!: • Stable Diffusion - Inp...
Background Replacement with Fooocus: • Stable Diffusion - Bac...
Fooocus Tips and Tricks (and AI Hands) • Overlooked Fooocus Tip...
Stable Diffusion - Styles in Fooocus - How do they work?: • Stable Diffusion - Sty...
FaceSwap and Consistent Character TipsPart 1: • Stable Diffusion - Fac...
FaceSwap and Consistent Character Tips - Part 2: • Stable Diffusion - Fac...
00:00 Intro
00:52 Where to get
01:45 Comparison Images

Наука

Опубликовано:

12 июл 2024

Ссылка:

Скачать:

Готовим ссылку...

Добавить в:

Мой плейлист

Посмотреть позже

Комментарии : 69

@KLEEBZTECH 3 месяца назад

If you find the video useful and would like to tip you can buy me some electricity for all these image generations. It is greatly appreciated! ko-fi.com/kleebztech

@michealwilliams472 3 месяца назад

Very cool video. The stormtrooper with his wife was actually super adorable.

@mrimpepe 3 месяца назад

"Gorgeous woman, 1970" AI creates a frame of Michelle Phillips. I agree.. Whom ever told you that, i agree.

@mrbabyhugh 3 месяца назад

2:55 as a descendant, this resonates with me. v8 did better ship and image composition but v9 delivered a sail that more suggest Viking, though the symbol isn't too distinct but it made an attempt at a symbol.

@ravenbandit 3 месяца назад

That calf quip was funny.. just the way you said it made me laugh so much.

@KLEEBZTECH 3 месяца назад

😂 I hadn't noticed it until then so it kind of caught me off guard.

@mrimpepe 3 месяца назад

Also, grand videos man, very good content walk-through and explanations. Direct to the point with some explanation and thoughts on it.

@KLEEBZTECH 3 месяца назад

Thanks!

@edward_fox 3 месяца назад

I haven't tried until I saw your video. For me, the results with my prompts got a lot better. Thanks for sharing the knowledge, personally now I don't know why that is not the standard one for Fooocus.

@KLEEBZTECH 3 месяца назад

It probably will be at some point. It hasn't been out that long.

@ThisClayman 3 месяца назад

Hey man, thank your taking the time to explain this :)

@KLEEBZTECH 3 месяца назад

No problem! A few people had asked about the difference so I figured I would throw something together.

@mrbabyhugh 3 месяца назад

3:11 v9. i can say this now. v9 seems to be better tham v8, but tge FOOOCUS v2 gives v8 some polish. Fooocus v2 wasn;t trained on v9 so I guess that why it doesnt polish it as well as it does v8. So matter of preference. If using Fooocus, use Jugga v9 without Fooocus v2 or Jugga v8 with Fooocus v2.

@KLEEBZTECH 3 месяца назад

I think for photo style images it is better which makes sense. The other stuff for the most part I think it is pretty close which is to be expected.

@megaerror007 3 месяца назад

very intresting compare thx

@KLEEBZTECH 3 месяца назад

You're welcome

@southcoastinventors6583 3 месяца назад

Do a video on Stable Cascade checkpoints and compare your prompts. Also a good trick is to do a image in several different art styles. Such as "(Artist), (Style of Art), A mystical sorceress, her eyes glowing with arcane power as she conjures a spell in her secluded mountain sanctuary." Then you can compare compositions of the styles feel free to omit either artist or style of art. Feel we need to see more checkpoint review in general.

@KLEEBZTECH 3 месяца назад

I'm not sure I can even run stable Cascade last I checked. And I'm not willing to upgrade my video card until I see what happens with sd3. That's the one I'm interested in most.

@Ramakrishhna 3 месяца назад

I think you can run Stable Cascade on Comfy UI. Coz even I am able to run it on my RTX 3060 6gb Mobile GPU. So try if its possible

@KLEEBZTECH 3 месяца назад

@@Ramakrishhna I'll have to check it out. Last I checked they hadn't gotten to that point.

@southcoastinventors6583 3 месяца назад

@@Ramakrishhna can also be run in forge

@southcoastinventors6583 3 месяца назад

@@KLEEBZTECH I was checking civit and there was a new checkpoint merge with playgroundv2.5 and sdx might want to test it out it works like a normal checkpoint

@techendofficial 3 месяца назад

Thank you bro❤

@KLEEBZTECH 3 месяца назад

Welcome 😊

@animatiV9 3 месяца назад

@@KLEEBZTECH Also thank you :)

@KLEEBZTECH 3 месяца назад

You are very welcome. 👍@@animatiV9

@mrbabyhugh 3 месяца назад

2:03 v9

@maga.rincon 3 месяца назад

I love your videos! Have you tried v9 with Face Swap? I could't make it work. The background is always blurry and doesn't respond to the prompt. Any suggestion?

@KLEEBZTECH 3 месяца назад

I have and just ran a few test runs and didn't notice any major difference. When you say blurry background I assume you are talking about how many photos are blurry in background when portrait type shots are taken? Words like portrait can make that worse. And since Faceswap tends to focus heavily on the face when you use it I think that has an impact as well. But overall I am not seeing a noticeable difference.

@maga.rincon 3 месяца назад

@@KLEEBZTECH Thank you for your time. I'm using a prompt like this: "Full body shot of a girl, [facial features]. Working out at the gym, wearing sneakers" or "At the gym" and the background is always light blue and nothing more. 10/10 images. Another one is "A girl with [facial features]. Soaking in the sand, beach, bikini, full body shot". The girl appears in the foreground and very far away the sea, but it looks flat, without grace. 10/10 too. If I use the same prompt in v8, the background looks amazing, but the girl isn't as accurate as in v9. Performance: Quality, Ratio: 832x1216

@KLEEBZTECH 3 месяца назад

First question I would have is does the reference image for the Faceswap have a simple light blue background? I know often that sort of thing can influence the generated image. Although I would expect using working out at the gym you would have some stuff that would look like a gym. Doing the same prompts without Faceswap do the images look fine? I would like to experiment more but right now without power and running on a generator after a big snowstorm so limited ability at the moment.

@maga.rincon 3 месяца назад

@@KLEEBZTECH Oh! I'm sorry to hear that you're facing such difficulties! I hope you're ok. You're right! It has a light blue background. I generated them with the prompt from one of your videos, but they came out light blue, not gray. Without Faceswap works fine! Thank you very much, I was freaking out! I'd try to remove the background.

@KLEEBZTECH 3 месяца назад

@@maga.rincon doesn't really matter what background it ends up with it's going to influence the final image. But often I overcome it by giving more details about the background like colors of walls or things like that. Now if you have multiple images with different backgrounds that actually seems to make it much easier.

@Peri2k 3 месяца назад

Nice! Is there any way to create in Fooocus two different characters in a photograph? for example a man shaking hands with Tom Cruise.

@KLEEBZTECH 3 месяца назад

Should be able to. Not on every attempt but on some. Another option is if both look like Tom Cruise change one after with inpainting.

@KLEEBZTECH 3 месяца назад

I just ran 8 images "an old man shaking hands with Tom Cruise" and got about a 50% rate of good images. Juggernaut v9. Just leaving it as man may not work as well.

@josemartins3514 3 месяца назад

I was wondering if you could clear my doubts. using a product (example box of chocolates) can I generate images where this box appears? I have been having difficulty with the letters on the packaging in question. Thx

@KLEEBZTECH 3 месяца назад

If you mean putting those in the image prompts and getting the same product, no. The only way we would probably be to train a lora or generate around the Box.

@josemartins3514 3 месяца назад

@@KLEEBZTECH thx

@diccionarioprimado 3 месяца назад

To define words, this works for me. example.... prompt: box of chocolates with the word written: "NESTLE"

@josemartins3514 3 месяца назад

@@diccionarioprimado I gonna try, but for prefect imagen i use Photoshop too

@diccionarioprimado 3 месяца назад

@@josemartins3514 Of course I would do it too, if you need a perfect final result... The idea is to alternate work tools like a good craftsman

@TheTchaume 3 месяца назад

There is a way to change the definition of the inpaint mode ? On a large scale of a large image, everything is blurry

@KLEEBZTECH 3 месяца назад

How large is the image? The models are designed to work at certain resolutions.

@TheTchaume 3 месяца назад

@@KLEEBZTECH thanks. images are 4K or more. With « vary mode » it seems to allow to work on such resolutions (no blurry texture when zooming) but not the same with inpaint. I don’t find in « developper » mode a way to improve resolution. It’s a pity not to let us try on bigger resolution (I’m ok to wait much longer for sharper result)

@mrbabyhugh 3 месяца назад

2:26 v9

@TheColonelJJ 3 месяца назад

Prompt ideas: I'm having one devil of a time getting an image with two distinctly different characters interacting with each other. Like a skeleton and a little girl with a detailed interaction. I may get both characters in one out of 15 images.

@maxtofly1606 3 месяца назад

Yeah it seems AI tecnology in not that far yet.

@KLEEBZTECH 3 месяца назад

That can be hit or miss that is for sure. Often you'll end up with two of the same. I think that's where the next version of stable diffusion will get better. The online ones have gotten to that point where they do pretty good. But I'll play around with your challenge tomorrow.

@KLEEBZTECH 3 месяца назад

So I just ran a few runs using prompts like "a skeleton and a little girl walking towards camera and holding hands" and got both in all images. Of course for some reason I get the skeleton walking towards the camera and the girl facing away. having her smile solved that. I was using the default settings and Juggernaut 9. And that was with the styles enabled. Are you having issues with longer prompts? Or even short ones like the one I used.

@KLEEBZTECH 3 месяца назад

I was actually very happy with some of the results. This was my final one. instagram.com/p/C5UGdgNOfRf/

@maxtofly1606 3 месяца назад

@@KLEEBZTECHNice, I noticed it works whne dealing with two very distinct characters, a person and an animal, or an sequeleton in this case. But lets say, "a girl wearing a blue dress, talking with a bald man" usually gets things mixed a lot, like both bald, or none, so I guess for more descriptive stuff, it is still getting there.

@fxlee888 3 месяца назад

@KLEEBZTECH Where is the parameter to specify the quality to 60 steps ?

@nenadm5747 3 месяца назад

I think it's automatic, 60 for quality, 30 for speed...

@KLEEBZTECH 3 месяца назад

Correct. Although they can be changed to anything in the debug settings.

@mrbabyhugh 3 месяца назад

2:46 v8

@MattyMilligan 3 месяца назад

Rum ham or milk steak? Hmmmm...

@KLEEBZTECH 3 месяца назад

😂 We know which one Charlie would want.

@fabiotgarcia2 3 месяца назад

My dream is if one day we could swap clothes, from real one that we upload.

@KLEEBZTECH 3 месяца назад

I don't think that will be too far off.

@user-bj5cp8qy2p 3 месяца назад

9 is much better and rich in fine details. Color tone can be adjusted easily but details are not. V9 is much improved. Annoy me a bit when you said you cannot decide.

@KLEEBZTECH 3 месяца назад

It doesn't annoy me if you don't agree. 😁