[New Feature] Train it the way that final image looks like a particular desired frame

[MavenDE]

Hello everyone!

I think it would be awesome if you could train final images with the goal to look like certain frame.

This would be much more flexible when for example the object moves or changes.
I am thinking about HyperNerf like solution, where if you open mouth during head/face image capture by smartphone,
it does not show anything distorted in the final render, but lets you pick which frame is the desired and final output frame.

The AI would basically try to find more details in other frames, but keep what it has in the selected frame.

Something like this hyperNerf:
https://www.youtube.com/watch?v=qzgdE_ghkaI

Anyone with knowledge good enough to update/fix this project to the level of HyperNerf?