Control AI Architecture with ControlNet

In this video we take a closer look at ControlNet. Architects and designers are seeking better control over the output of their AI generated images, and this neural network process provides a way to do that.
The process uses a variety of techniques to lock the image composition allowing you to apply your prompt directly to the subject or geometry within your image. Some of techniques it uses to achieve this are Canny Edges, which track either prominent or distinct edges in your image, or depth maps which qualify spatial attributes to your image.
Stable Diffusion ControlNet:
stablediffusionweb.com/Contro...
ControlNet Scribble:
replicate.com/jagilley/contro...
Scribble Diffusion:
scribblediffusion.com/
00:00 - Intro
00:34 - Stable Diffusion (ControlNet)
02:43 - ControlNet Scribble
05:57 - Scribble Diffusion
10:08 - Final Thoughts
#controlnet #artificialintelligence #architecture

[@SebAnt]

Thank you Stephen for these wonderfully informative videos you put together 🙏🏽

[@StephenCoorlas]

My pleasure! Thanks for your support 🤘🏼

[@monikasaringer1152]

That's great! I love AI, I know there is alot of criticism but there is no doubt at all that AI is only going forward, there is no turning back now & it is probably going to be a part of our everyday lives in the very near future

[@StephenCoorlas]

I agree. I believe it’s better to embrace it and be productive about finding ways to responsibly integrate it within our professional and personal lives 🤘🏼

[@knight32d]

Wow! 👍👍👏👏 You're the man. I'm so happy to see the Scribble Diffusion option. Is it part or separate from the ControlNet app?

[@StephenCoorlas]

Thank you! Scribble Diffusion is separate, I have links in the video description to each app/website. I do believe scribble Diffusion uses ControlNet behind the scenes to lock the image geometry as well.

[@atmaatma6867]

Thank Stephen , can we turn sketchup image in ( black and white ) to Stable Diffusion and Prompt to make the Rendering of those Image in S.D .Please tell me how's work .many Thanks

[@StephenCoorlas]

Yes you can, simply export your SketchUp image and upload it to Stable Diffusion’s ControlNet 🤘🏼

[@MyAI6659]

Hi Steven, Thanks for making these awesome videos. Could you please make videos about how to train Stable Diffusion models on architectural images? And also give us info about which SD models work better for generating architectural images? Thanks a lot brother.

[@StephenCoorlas]

Hey - I haven't actually trained any SD models myself. I've been hesitant because I know of several companies who are developing programs/software that will make it even easier to train models by uploading your own image sets and assigning labels/attributes to the different images. I will be doing a video on that program hopefully shortly. Thanks for your support!

[@MyAI6659]

@@StephenCoorlas That's great news, Thank you so much. 😘

[@joojaa3927]

You would probably get better results, if you trained an image on a few dozen existing images first. I mean the stable diffusion base model is quite bad anyway. But with a few rounds of lora training it can be just as good if not better than some of the other tools.

[@StephenCoorlas]

Thanks I'll need to look into that!

[@almirakvic7791]

Nice. Is there solution to upload rendered user's 3d model with complete materialization or studio photography and then change environments like rooms setups like in Vray, Corona etc ?

[@StephenCoorlas]

I've been in discussion with someone developing a process/software like that, but I'm not aware of anything that comprehensive that exists right now.

[@islandersean2213]

Can you do a video ai floor plan generator. Thank you 😊

[@StephenCoorlas]

Hey - Yeah I'll be getting into some more real applications of these technologies in my practice. Thanks for tuning in!