Видео

I appreciate your suggestion! I actually have a german Channel that was stuck on 80 subscribers when this Channel hit 2K subs so i focused on the englisch channel. KIWissen2Go. Currently in the planning of releasing there again...

Hi, I am not sure if I understand your question correctly, but since Pony is derived from SDXL i would use the 4xLSDIR Upscaler. Happy creating.

​@@AIKnowledge2GoI ask for the same guide but on sdxl/ponysdxl if it possible

and thank you so much for upscaler

That is a little bit tricky, inpainting would be the easier solution here.

nah been trying that for month in supposed latent mode that also broke with forge inpainting them is probably beter. you may wanna look at Krita ai diffusion for that

I am using Darkreader in Firefox, but i have also set my windows to darkmode by default. So most apps start in darkmode if supported. Another way ist to customize your webui-user.bat file in your stable diffusion a1111 folder. change this: set COMMANDLINE_ARGS=--theme dark You need to start a1111 via the webui-user.bat file from now on. Happy creating.

@@AIKnowledge2Go I also use dark mode but Darkreader is what I needed! Thank you 🤝

@@nenadm5747 you are welcome.

I still have Krita on my list for software to try out. Maybe a good topic for another video. Happy creating.

YES EXACTLY MY ISSUE, i'm two week into Krita AI and it have what i want most of the time but region prompt is not there. (workaround i possible and comfortable at least) i loved regionnal prompter since day one but A1111 is a NO GO for SDXL even with my 3090 (VAE decode is awful above 1536*1536 pixel) --- ForgeUI output way faster but it broke the extension in some ways SDNEXT i still have to try it since the recent UI overall rework. it's also way beter than A1111. ComfyUI have his own beter way of doing region attention but is weird to use.

thats what i want to loook in too.. otherwise, i use Invoke Ai, also as it seems following a by far more intuitive user interface.. here also you can draw regions and use regional prompting, and you also can combine IP Adapters with those regions. which realy gives you even more control. Aside from that the unified canvas can give you even more control.. well but Krita also sounds good.

@@tomschuelke7955 i really need to go back to invoke as it tries really hard here and now. canvas was always a big forte of invoke. if Lora regional work i'm sold. if not i found that there is SD Forge Attention Couple extension that i could use for that. (two SD at the same time ouch... Krita run background ComfyUI) Krita is great but won't do region yet.

What at the moment works are IP Adapters for different regions. But it s very intuitive. different Loras at the moment are not possible, because they fit into the overal Ai model so influence the whole image... never the less i used this for telling architectural images, that have to reflect exactly your CAD screenshot, which materiels to use where in the image... its not yet perfect, but allready a biiig step in the right direction.. For architectureal images also you have to say, theres no AI model trained anywhere, that is good enough... For art and sketches and free images never the less this regional prompting is very powerful ruclips.net/video/CLVylJAMIF8/видео.htmlsi=oemKX7zqDANyjWof&t=93

You're welcome!

"sis" and "set" could also use some improvements - it's not "set" hard to put some effort in pronunciation. A good teacher would insist in such "sings", but german teachers suffer from the same problem.

Ah, the rogue "rogue" strikes again! My spelling could indeed need some work. I'm glad you enjoyed the video

Your support means a lot to me, thank you!

Ich finde ich hab mich neueren Videos deutlich verbessert.😂 Aber dieses Video war schon schlimm was den Akzent angeht. Danke für das Feedback.

Thanks for the heads up on resolution tricks! You are right.

Hey! Thanks for the love and the suggestion. I'll see what I can do for the SDXL video you're looking for!

@@AIKnowledge2Go That would be great! I'm thinking SDXL isn't worth the time at the moment because of the lack of tools that work well with it... Do you have a discord or any other way to contact you besides your paid Patreon account?

Have you already watched this one here? Another tool video is definitely on my list. ruclips.net/video/8At23HEJnAk/видео.html

Toonify is great, you are right. At the moment I am experimenting with Level4 XL.

You're welcome

It seems like the IP Face Adapter is becoming the go-to solution now! I will make a part II video of tools for A1111 soon

@@AIKnowledge2Golove your content. Literally perfect for me. I know enough to where these details in what you cover perfectly enhance my work. Please please please cover SDXL and what control net features actually work. IPFace Adapter seems to not work on SDXL and prompting and score values on XL and Pony etc make no sense to me:)

ja ohne Probleme, aber ich empfehle dir auf die neuste version 1.9.3 (derzeit) zu upgraden.

@@AIKnowledge2Go danke dir, hatte bis jetzt nur fooocus aber denke mit automatic wird es vielfältiger

Ich könnte aber auch Schweitzer oder Österreicher sein... 😂

Personally, I prefer AI generated images more than drawings. They have more quality and can be easily adapted to whatever style you want to. But I know you're talking about the effort. It's true that doing AI images needs less effort than painting or drawing, but it has a process you must follow and can take you even 6 hours to do a decent image. It's not like you ask to the model and it will give you all, perfectly done. No no, it's a long and boring process using Photoshop (for example), creating the best prompt from hundreds of ways to describe the same thing, selecting the correct model for that specific part you want. Is it easier than painting and drawing? Yes it is. Does that mean you can't be an Ai Artist? Nope. Doing more effort does not make a better artist, get that wrong idea out of your head.

Thank you for your comment! I understand that the term "AI artist" can be controversial. Perhaps "AI art creator" or "AI art conductor" might be more fitting. The field is evolving, and so is the language we use to describe it. I'm open to suggestions and happy to discuss further!

I appreciate your kind words! Another workflow will be my next project.

@@AIKnowledge2Go thank you!

Glad it was helpful! Happy creating

The VAE is basically an Image Enhancer. When SDXL came out it needed a refiner. It doesn't anymore. So you can just keep using your SDXL VAE. Just don't mix SDXL and SD 1.5 VAEs or you will get strange colors.

My guess is you are using Automatic 1111 version 1.9.3 right? Just try the following. Got to extensions Tab and click "Check for Update" then "apply and restart UI" and controlNet should work again. update 1.9.3 broke controlNet but they fixed it.

Thank you for the positive feedback! Happy creating.

thank you so much I am really struggling to get good results like everyone else seems to have

I actually will make a video about in the near future, but don't expect wonders from it, apart from slight movements and maybe a little camera Turn. Happy creating

@@AIKnowledge2Go I don’t get why it’s so hard to create moving images? We can create a frame that’s amazing .. even in quality .. but in motion cohesive and stability breaks down

​@@AIKnowledge2Gomaybe try using ipadapter and faceidplusv2 to generate image base on your own portrait then add a prompt for movement then send it to animatediff?

Thank you for your kind words and subscribing. There's actually an improved version of this workflow you can find in this video ruclips.net/video/wyDRHRuHbAU/видео.html it uses control net so it's a little more advanced. Happy creating

Ha da hast du mich erwischt 😂 Deutscher durch und durch😂 Du hu hast absolut recht, das Video richtet sich auch eher an Leute die bereits Berührungspunkte mit Automatic 1111 hatten. Ich kann dir dieses Video von mir empfehlen da erkläre ich die einzelnen Settings: ruclips.net/video/SHikMK39Q30/видео.html Wenn du der Liste folgst da hab ich einzelne Themenbereiche in Reihenfolge abgedeckt. Ich hoffe das hilft dir weiter.

If you are using the latest version automatic 1111 this is now a new dropdown it's called scheduled type you can leave it as automatic, but you can also set it to karas by hand. I hope that helps happy creating

@@AIKnowledge2Go thank you!

Aw, thanks!

I'm sorry to hear you have problems with in painting. Try to alter your prompt just write about the content to be in painted. Another thing you could try is lower the denoising strength further. I recommend you to watch my after detailer video. It's a great tool for automatic 1111 that in paints faces automatically ruclips.net/video/y3DxX9s0NhQ/видео.html I hope this helps. happy creating

I haven't tested if that LoRA works well with realistic checkpoints but you can try to use for example cyber realistic or juggernaut from civitAi.

Regional prompting is definitely a brilliant suggestion for an upcoming video, and your interest has made me consider it more seriously. I often use inpainting and outpainting in Stable Diffusion to address minor misplacements, but haven’t explored regional prompting in depth yet. Your idea has definitely inspired me to dive into this topic for future content.

@@AIKnowledge2Go Will look forward to it! I think it's very under utilized because it can be a little challenging to understand how to properly set up the divide ratios and build the prompt. You might look into one or more of these extensions if you do a video on it: Regional Prompter, Latent Couple, Composable Lora, Lora Block Weight, & Dump Unet. Can do some interesting things with various combinations of those 😁