SDXL 1.0 in ComfyUI - Stable Diffusion

preview_player
Показать описание
How to get SDXL running in ComfyUI. Updated with 1.0 links.

Рекомендации по теме
Комментарии
Автор

Just a quick tip, instead of disconnecting nodes, you can just select them and press Control-M to disable them... or 'Mute' them. Same shortcut to turn them back on.

franckyb
Автор

Thanks for these videos! Really helps make things easier

sarundayo
Автор

I tried it and works great. However, the Upscaler is not working. Any solution? Thanks.

consigiere
Автор

If you think you are never going to use the non refined images, you can make a generation more efficient by removing the vae decode and preview output for the base model, it adds up over time and with higher batch sizes.

jibcot
Автор

Thanks for this video, tried this and loved the result. Would love to know a lot more about advanced work of stable diffusion on comfyUI( Regarding Control net and upscaling)

musicandhappinessbyjo
Автор

I spent a LOT of time learning SD and all of it's features over the last months. Became quite good at it thanks to you and a few others that explain things so well. I'm on the fence with this ComfyUI thing. First, it feels like starting over and learning it all....again. Second, it just looks convoluted and confusing and not streamlined at all (yet). Every time I see one of these video the "workflow" is all over the place and I can't help but think that there just has to be a better, simpler way to have a UI for this thing

theguru
Автор

He needs to give up dad jokes for good…

… and only tell dad jokes for evil.

ayrengreber
Автор

How can you do img2img, can you make a video? (without having an image+the workflow in it)

laurent
Автор

Can I have the link to the image or directly the 2nd workflow? Thanks

dviethere
Автор

can you explain why there are sizes of 4096 in many of the settings there?

WillakaPlaeground
Автор

My version of comfy doesn't have any upscalers showing in the upscaler box on your workflow. Which directory should they be in so that it recognizes them?

ThoughtFission
Автор

Hey, Sebastian! Can you help me out with something? No matter what I do the depth midas preprocessor in comfy ui wont work. I get an error that there is no midas. I tried everything and am wondering if this is a me problem, or a current install issue.

beatemero
Автор

I'd love to see more videos from you for comfy, especially for controlnet usage and how you would approach inpainting as this seems to be not es straight forward as in 1111. Thank you

DanielPartzsch
Автор

hello! i hope you do a guide on using loras in comfyui through stable diffusion 1.5 or a1111

sonata
Автор

Mine gets stuck in the "Load Upscale Model" and never generated the upscaled image and the node gets a red outline.

Failed to validate prompt for output 190:
* UpscaleModelLoader 192:
- Value not in list: model_name: '4x_NMKD-Siax_200k.pth' not in []
Output will be ignored

JDRos
Автор

Hi @sebastiankamph, can you share the updated json file of the workflow for SDXL 1.0. Thanks

victorrinoalonzo
Автор

To auto generate the workflow do the images need to be generated with comfy UI? or we just drop any AI image there?

arannak
Автор

Do you think with a 1080 TI should try SDXL or better don't waste time trying?

ZeroCool
Автор

I can see how the node system can confuse some but it's actually pretty good for those who want pixel perfect control. They should make a simplified version and make the node version optional.

NineSeptims
Автор

There is a ComfyUI extension for Automatic1111. Is there any advantage in creating a separate install like this over just using the extension?

Aegisnir