>But if costs are an issue, the system is pretty easy to improve in theory: if you keep the dialogue softsubbed, only the parts of the video that actually feature typesetting would be hardsubbed, and with some clever engineering and an understanding of how modern media formats work, you would only have to keep multiple copies of the typeset parts. And since the average anime episode has on-screen text only for a small percentage of its total runtime, combining softsubbed dialogue and hardsubbed typesetting like this would make for a highly cost-effective setup.
Not sure why the article keeps wasting time with this idiotic mixed system since hardcoding multiple transcodes for 100+ languages/formats just for small typesetting differences is obviously not an option for either Netflix nor Amazon.
Resolving this issue properly everywhere requires changes in the current standards being used.