Honestly so far I'm not sold on it. On paper it seems amazing and in lots of ways it is, but practically speaking for me it's not a game changer... yet. Being limited to odd crop sizes and limited frame rates (i.e. hard to get decent res at 60fps) is not great. Fully understand the hardware limitations all the ML contributors have to work with. The effort and ingenuity being put in to work around the limitations is amazing but I need more time with it to see if the downsides and extra workflow elements is worthwhile for me. So far it seems like if the scene is exposed properly there isn't a massive noticeable difference in the final footage with raw vs the default h.264. But again I need more time to fine tune the settings and figure out the best workflow.