TouchDesigner Walkthru - 360° Projection of AI-Diffused Content
HTML-код
- Опубликовано: 15 сен 2024
- This demonstration is an approach I used when first attempting to map images created from AI onto a spherical projection. Many different methods have arisen since the uploading of this tutorial, however I hope to inspire and share my first techniques to those attempting their own design. The process includes using a commercial license of touchdesigner in order to enable high resolution images needed for projection. I encourage enthusiasts to improve upon this concept and to find their own more efficient methods of achieving the same.
Tools & Resources used:
derivative.ca/
polyhaven.com/...
Thank you. This is a very good starting point to go into AI VR video.
I'm amazed by AI, but I'm also amazed by humanity. There's so much brainpower showing in this.
I legit share your videos almost daily! Super inspiring. What you're doing is so fascinating. Between you and and the riffusion devs my mind is blown.
Very cool! Thank you so much for this tutorial!
Thanks for sharing Scottie!
Awesome. I'm going to do this to my office.
Amazing tutorial. Thank you so much for sharing!
Nice video, Thanks.
Beast mode. Thanks.
Owww man cant wait to try this!
Great stuff ❤️
great work! how would you create a click and drag implementation like youtube’s 360 viewer?
Amazing! Thanks for sharing :)
Amazing work. I was really curious you talk a bit at the end of the video that it would be possible to make the diffusion be automated. I've been experimenting but no luck. I was tried using the api (from Chat gpt) from a text-to-image Ai and then constantly sending the final image back to recalculate some other parts of the image? So I would have a constantly changing image. I couldn't found a way to diffuse smaller parts with a percentage or something so I added a noise map with changing seed for the parts that would be recalculated. Could you please explain a bit further how it is done?
Sorry if I missed something, but what step creates the "morphing" between the original and diffused image?
omg do we have access to any of this to play with yet!?
Howdy! Any plans to make more in depth tuts regarding that topic? Great work! Cheers!
cool
hello, sometimes i don't understand some parts because english is not my first language, and particulary the one with multiple images. Anyway, this seems to work too with the base image only. Great tutorial thanks.
Have you tried connecting TD with an VR viewer such as oculus to view the content?
This is amazing! I've been following your work for a while and was very curious to see your approach in TD. I have a question - how do you diffuse the image in MJ or another SD model without too many changes? I've been trying for a couple of hours with different prompts, no prompts, image weights etc. Everything I try I get a completely different image from my original HDRI / equi. Is there a command to ensure that the original image stays (mostly) intact while adding diffusion on top? Thanks for your time!
If your selected AI image generator has a % option or a "strength" option, try values of 50% or 0.5, then adjust if needed. It also depends in some cases which sampler you're using, as well as number of steps.
@@scottiefox2525 Thanks for getting back to me so quickly! I’ll try that out
amazing, damn resolution limit makes it pointless. still very interesting! shame they do not recognise autodidacts as students!
how do you exporting the render from touchdesigner to get a 360°degree file that i can upload on youtube?
That's a completely different workflow that includes exporting an equirectangular video with a 2:1 aspect ratio using Premiere Pro set to VR mode during export. RUclips interprets that footage as 360⁰ and uploads accordingly. You can search for tutorials on that subject for more info.
is this rendered in real time or just a rendered video?
This is actual output of using the software. Touchdesigner is a platform designed around real-time visual content.
You are awesome. You need 'promo sm'!!!