Unforunately Youtube update their methods all the time. The latest known guide can be found here.
Essentially, you should save the videos as Parallel MP4 files. Then download and install FFMPEG.
Finally transcode the video and add the required information using the following command:
ffmpeg -i input_file.mp4 -vcodec libx264 -x264opts \ "frame-packing=3:frame-packing-interpret=1:frame-packing-quincunx=0:frame-packing-grid=0,0,0,0" output_file.mp4
Where 'input_file.mp4' is the file you saved from 3DCombine.
If your nVidia GPU supports hardware H265 video encoding it will be enabled automatically when saving compressed AVI or MP4 files. You can check for this in the about box. It will say 'hardware video compression available.'
Certain nVidia GPUs can be used to accelerate the intellgent 2D->3D conversion mode in 3DCombine. Known supported GPUs are RTX2080, known unsupported GPUs are RTX3080. This should be considered an advanced user operation. These instructions are valid for 3DCombine v6.17.
Download and install CUDA 11.0 (to get the right version select Windows, x86_64, 10)
Download and install CUDNN (free nVidia sign up required)
Download the Tensorflow 2.4.0 GPU archive and replace the tensorflow.dll file in the 3DCombine installation folder.
Open 3DCombine and generate an intelligent depthmap. Open the about box. If successful this will report 'GPU acceleration available'. Note that this will disable usage of the GPU for video compression when active.