why compressed video looks blocky, and how to fix it.

Divide the video bitrate by width x height x frame rate and you get bits per pixel: how much information the encoder can spend on each pixel of each frame. When it falls too low, the encoder starts describing whole blocks of pixels at once, and you see them.

Basketweaver treats 0.03 bits per pixel at 30 fps as the floor for a watchable 16:9 picture. That puts a minimum bitrate on each resolution:

If your bitrate is below the floor for your resolution, step down a row. A sharp 480p picture looks better on a phone screen than a broken-up 720p one.

At a fixed file size, doubling the length halves the bitrate. This is the resolution the converter picks at a 4 MB target, from a 1080p source, for each clip length:

• Lower the resolution until the bitrate clears the floor in the table above.
• Trim dead time from the start and end. The bits go to the seconds you keep.
• Drop 60 fps footage to 30 fps. Half the frames, twice the bits per frame.
• Drop the audio if the clip works without it.
• Use a two-pass encode so the encoder can save bits on still scenes and spend them on motion.

Grain, rain, confetti, fast camera pans and screen shake all change most of the frame every frame, which leaves the encoder little to reuse. These clips need more bits than a static shot of the same length. A light denoise before encoding gives the encoder less noise to preserve:

ffmpeg -i input.mp4 -vf hqdn3d,scale=-2:480 ...

Screen recordings are the opposite: large flat areas that barely change. They compress well, but small text blurs first, so keep their resolution up and cut the frame rate instead.