This is a cool approach to improving the speed of training and rendering but seems to come with some limitations when it comes to representing reflections.<p>It reminds me of some of the experiments and approaches used for the custom rendering approach used by the game Dreams:<p><a href="https://www.mediamolecule.com/blog/article/alex_at_umbra_ignite_2015_learning_from_failure_video" rel="nofollow noreferrer">https://www.mediamolecule.com/blog/article/alex_at_umbra_ign...</a>