I don’t use OBS to create the videos, I make them by hand. I record the VR output using Nvidia Shadowplay, this is where I get the main race feed. I then record a video of the replay separately for the picture-in-picture part.
I drop the videos into Davinci Resolve then drag them onto the Edit timeline. To make the PiP I zoom the replay in and position it in the top corner using the properties.
To sync the videos up is a pain, you have to look for something that happens in both, such as crossing a white line, or a good tip @Brave gave me once was to look for a bridge and use the shadow in the cockpit. There’s a knack to it and you get better with experience.
I then go to the Deliver tab and render the video (mp4, network optimised, medium quality, high baseline).
The race chat is a pain. I have been doing a lot of Missed Apex stuff recently and the Discord chat is included in the VR audio, which obviously I can’t use. To fix this I make a second video of the replay from the cockpit view and then use the audio from that.
There are tools you can use to separate out the audio tracks virtually but I’ve not tried them yet and apparently they can have a big performance hit.