Leaving this here, in case anyone’s wondering about the large file sizes for these uploads. From the encoder’s perspective, there’s so much happening on the screen that it simply can’t be stored very efficiently. There are constant small changes in the image, and flat areas are almost nowhere to be seen, due to the combination of grain and the analog coloring approach of the film. Not to mention the snow particle effects in some scenes. The background animation sequences are the worst about this, since practically the entirety of the view is changing.

I’ve tried to keep these uploads as optimized as possible without ruining the quality by choosing a sufficient x264 constant rate factor (20) and the slowest reasonable preset (veryslow). Maybe one could be able squeeze them into a smaller size by doing some appropriate filtering before encoding, but I wasn’t able to achieve this with meaningful results.