Відео

Hi danke für das Feedback. ich habe tatsächlich Pläne zu sowohl inpainting und auch outpainting neue videos zu erstellen. Leider kann ich noch keinen genauen Termin dazu nennen.

No, its a completely different installation. Your A1111 remains untouched.

Glad you enjoyed it!

You can train both in a single lora, this requires a more delicate image selection and the rendering of a lot more images. If the body has to be of nude form, then good luck. You may want to use a textured 3D Model for this and render these Images in a 3D Environment, like Blender or Unreal Engine 5. Hm... actually not a bad idea for a new video... *taking notes*. You also could take real photos, but make sure you have written consent from the person to use them.

Thanks for you feedback. Since most of of my viewers have an RTX 2080 or lower ( i did several polls on this). i decided against local training. I am using the Kohya method in Civitai online trainer. This can be applied 1 - 1 to local kohya and should yield similar results.

By including images of facial expressions into the dataset (for sd 1.5 and SDXL there are facial expression loras), you can train your character lora on these. Originally i planned to show this in the video as well and include it, but the video itself is already 19 Minutes long so i got rid of that in the video script.

Hi thanks for your comment. Regarding the VRAM. The VRAM limits only your capability which base model your pc can handle. With 6 i recommend to stick to Stable Diffusion 1.5 Models. SDXL or FLUX will probable be to much for your Graphics Card.

mimicpc's tutorials are very nice explained, recommend for newcomers

@@Bpmf-g3u thanks, I ended up spending the day repeat inpainting until I got it decent. Tried a few other things but none worked at all for some reason.

When doing Tile Upscale its important to turn of restor faces, because its going tile by tile that's why you need to turn it off. Regarding your problem usually you want to fix all your inpainting at lower resolutions and then upscale, if anything in the image distorted then decrease the denoising strength.

@@AIKnowledge2Go funny enough it looked fine before the upscale, perhaps I needed to change the denoising. I’ve noticed I added TOO much detail after the upscale and tiling. Would lowering or increasing the denoisong fix that? Sorry for all the questions I’ve never touched the upscaling before so this is all new to me. Appreciate the response!!

Skipping layers can lead to variations in the generated images, potentially yielding results that are more aligned with your creative intent. Hope that helps.

Hey there!

I wouldn't directly compare them, since is SwarmUI has "just" a comfyUI integration and Stabillity Matrix can handle a lot of different UIs. When it comes to image rendering with complex workflows i use comfyUI, for Single image purpose where i know i have to do a lot of Inpainting and moving back and forth i use Automatic 1111 because some extensions aren't available in Forge yet.

Hi its hard to keep up these days. I'll make a video about it.

Hi yes it is. Maybe you want to use controlNet for this. Here is a newer version of this video. ua-cam.com/video/wyDRHRuHbAU/v-deo.html To remove objects just write what you want to have instead on the background for prompt when inpainting. You need to experiment with the denoising strength.

I feel you, currently we are somewhat stuck with the ai models...

@@AIKnowledge2Go Yep :(

Ask the guys at Stabillity AI or ComfyUI developers (Maybe they changed it in the meantime, i don't check small issues). Happy creating.

I haven't checked recently, but maybe its time i do...

Hi, if you tell me where you are stuck, maybe i can help. So it does not lunch when you double click the exe file?

@@AIKnowledge2Go my bad I was down loading the wrong thing

I'm so glad to hear that the inpainting function has made such a difference for you! It's always rewarding to discover new features that enhance our creative process. Keep experimenting! Thanks for the remarks on my accent. 😊

From what i heard some loras work with nf4 others don't. I am using loras in ComfyUI.

When you have a newer nvidia card you can try the studio driver instead of the "game ready" driver. For AMD i don't know. Does this happen only with flux, does this happen also in ComfyUI?

Yup (suffering)

You're welcome! Always here to help out.

Welcome to the world of comfyUI! Thanks for the remark. This is all changing so fast, i will surly make an updated video anytime soon. Happy creating.

Haven't used it in Forge tbh. normally you put the unified model in the Models\ControlNet folder, then open the ControlNet dropdown, Click the correct preprocessor, Canny is for outlines, open pose for pose etc. upload your input image, put in your prompt and in theory it should work.

Ah the 1050 was a great card at that time. I admit for Flux it maybe a little bit low on VRAM. Have you tried online tools like Civitai?

@@AIKnowledge2Go Yea its better to go with online services for now until someone can crack it to work with those low 4GB VRAM. Tbh most of old potato rigs will still have that amount of vram right now. Its still working good for all of those years its kinda sad to just buy a new one 😁

Nope the 3090 is great actually. 🙂

lol I'm on a 3060 and getting an out of memory error message, so I guess so.

Since the models get loaded directly into the VRAM of your graphics card that is not possible unfortunately.

What happened?

@@AIKnowledge2Go It looks like the same error message as in the video, but no update is available..

@@MrDefix Sorry to hear that. Then try using a different branch, when installing click on branch and select 2.0.780 for example

@@AIKnowledge2Go Big thanks, that's helped me.

wow 4 hours for a single image, that I call dedication. For a 16:9 Image I start with 1280 x 720 and then I upscale. Maybe this approach can work for you as well.

@AIKnowledge2Go yeah I male smaller and adjust in photoshop, I just change the dpi from 72 to 300 omg it's better and takes seconds.

Sounds like you did a 3d render 🤣 Actually nah with a 2070 a 2k image with the right setup in render settings within blender cycles would probably take you less than 5mins

Lol my 8gb gpu fails when i do controlnet

8 - 12 if you are still having trouble you can try the flux1-dev-fp8.safetensors file i liked as "fallback" in the description.

i hot 6vram GTX970 takes 2 min for 1. 5k image

I'm glad to hear you're willing to give Stability Matrix a try! It really makes a difference if you are using multiple UI's.

Thanks for watching! I'm glad you enjoyed the video!

Yes, nf4 is already obsolete. :-)

Yeah i have noticed gguf. The development is quite fast in terms of flux. I could make videos 24/7 currently.

Please do, we love the content 😉 ​@@AIKnowledge2Go

whats the gguf models? Can i run it on my 3060TI? cause i'm currently using the nf4 like in the video

I'm really glad you found the tutorial helpful! Your support means a lot to me.

@@AIKnowledge2Go Well deserved ;)

Yes!!! WOOHOO!!! *BJ* *HJ* *RJ* lmfao

Vielleicht sind die noch auf Malle? 😂

@@AIKnowledge2Go danke für das video bruder xD

I'm glad to hear that Control Net is helping you out! It's amazing how a little tool can make such a big difference in your workflow.

You would first have to outpaint the image (into wide screen) and then send it to inpaint and mask the area where the character you want to add should stand and use a high denoising strength, 0.7 - 0.9. Use ControlNet inpaint with either llama or global harmonious as preprocessor. Remove everything from the prompt that has nothing to do with render. "Phototrealistic, 4K, HDR" etc. has to stay in order to give you the same look. Working with Character loras here can be tricky. I have a guide on my patreon in my 3 Dollar tier, but here is a free version that you can get you started: www.patreon.com/posts/99183367 Hope that helps.

@@AIKnowledge2Go I have another question. When putting the denoising strength to 90 for the image detail increase, I sometimes see changes happen to the image that I didn't want to change from the original image, but it's random. What is a good way to keep the image 100% the same, but still add the detail you did in this video?

I terms of tooling, SD 1.5 is still the best! The SDXL ControlNet Models for example are sh*t. Inpainting, outpainting etc. just worked in 1.5. Happy creating.

Can you use ComfyUI with any other workflow, lets say a basic SD 1.5 or SDXL workflow? I don't know much about your system so i may be wrong here, but to me that sounds like your system tries to render using the cpu or an onboard gpu.

@@AIKnowledge2Go my cpu doesn't take too much load during the process and gpu takes around 15% and my ram (32gb) starts to fill up and goes upto 31.6 and remains there until i closs the dos window to restart the comfyui because of too much time passing. Yes i am using and can use sd 1.5 and sdxl but on my flux workflow i am only using flux. Only flux is not working for me. I tried both full model and the schnell model and getting same issue. My system is I9 13900k Rtx4090 32gb 6600MT ddr5 ram

@@zaselimgamingvideos To be honest, at this point, it feels a bit like we're grasping at straws, but let’s try a few more things to diagnose the issue: Have you tried setting your model weight to FP8_e4m3fn and configuring your NVIDIA card to prefer no system fallback? Sometimes these settings can influence how resources are managed. Can you confirm if CUDA 1.21 is installed on your system? Having the right version can be crucial for compatibility and performance. Additionally, a fresh reinstall of ComfyUI might help clear up any configuration issues or corruptions that might have occurred during the initial installation or during use.

@@AIKnowledge2Go yes my nvidia is previously set to no fall back. If i remember correctly my cuda says 12.something when i load automatic 1111 or comfyui. Actually it is a fresh install, i did after your and couple other videos. Before that i was using A1111. I installed comfyui to try out flux. And yes i tried both default and fp8_e4m3fn just to check if comfy is not registering my vram as 24gb. But it does when i launch comfy, bram 24gb appears on the doss window. I will check confirm the cuda version on comfy but i am 100% its 12.something on my a1111 after the recent update.

@@AIKnowledge2Go i just checked my cuda version by command nvidia-smi and it says i have cuda 12.5

Haha, glad you enjoyed that moment! It definitely caught me off guard too when i wrote it 😂!

@@AIKnowledge2Go - Nice explanation, BTW. - Thx for your effort. - I started in computer graphics, and 3D 30+ years ago, and the progress, especially recent, has been amazing and fun to watch (and be part of: as a modeler/animator, and developer of custom code for 3D). - Keep up the great content!...