Видео

That sounds cool. It might work as a way of using Android phones as input to our open video nerf engine

@@lifecast7823 Did you see the water balloon burst in the 2019 paper ? « Wireless Software Synchronization of Multiple Distributed Cameras » This is insane 🤯

@@thereisnomouse Cool!

QR code alone is only good to about 1/10 sec in my tests. We'll have to use a few other techniques as well

a couple of meters I think

NeRF requires many (30+) images of an unmoving scene from all different angles to create a model that can later be viewed from any angle. Baseline NeRF doesn't handle video or scenes where anything moves. NeRF is also not fast enough to render on mobile devices currently, and not easy to edit with existing video tools. In contrast, Lifecast's software needs only one or two images, can do video, can be edited, and can render on mobile devices, but this comes with a tradeoff that the 3D reconstruction only looks perfect when the virtual camera is exactly on top of the original camera, and the farther away you move, the more artifacts appear. However, we are always working to improve our reconstruction quality, which is powered by machine learning and 3D geometric computer vision.

Please contact info@lifecastvr.com

You can use an iPhone that has LiDAR. There’s a plug-in for blender and you can play the scene in 3D while you move around it. It’s not great .. but it’s good. I think you would need a lot of gear to be able to have a full 360.. or maybe something down to 3 or 4 cameras with LiDAR.

Sorry for the confusion, we updated the software to improve visual quality. Please download an updated Unity example project here: drive.google.com/file/d/197Ea3MHUKMsS4BUy86iVwGukmClV0V_9/view?usp=sharing

@@lifecast7823 what about the web player? same issue

it's just for the background, nothing is supposed to be interacting with it.

@@DimiArt consider changing the video title then: Stable Diffusion to 3D Unity background art in 2 minutes

We don’t use Midas for vr180 input. It’s a stereo depth estimation method using deep learning. We use the distance between the two cameras to produce results in units of meters, accurate to within the limits of calibration and disparity error.

We are working on something like this. Stay tuned!

@@lifecast7823 Nice looks really cool so far, i have been really wanting to dive into this and understand how I can tweak it and utilize this even more. Any suggestions of where I should start?

@@xecense lifecastvr.com/tutorial.html

@@lifecast7823 Yoooo thanks brother !

@@lifecast7823 Thank vm bro ! but please wich parameter do you use to generate the initial image? wich dimensions etc?