A talking head in a vertical frame, with the speech shown as huge captions, one to three words at a time. Each word pops in at the beat, and the current word takes a bright colour.
Core thesis
Make every word visible at the moment it is spoken: the caption is the main picture, and the viewer reads and hears the same beat.
Design principles
Set captions at 72px to 110px on a 1080x1920 frame, with 1 to 3 words on screen at a time.
Place the caption block at 55% to 65% of the frame height, centred, inside a 64px safe margin.
Colour the spoken word in yellow (#FFE600) or green (#3DFF7A) and keep the other words white.
Add a 6px black outline or a 4px hard shadow to each word.
Pop each word in at 90% to 110% scale over 6 to 8 frames at 30 fps.
Time each word to the audio within 2 frames.
Use a bold, heavy sans in capitals and one emoji or icon per sentence at most.
Anti-patterns
Full sentences of 12 words on screen
Thin, serif or script type
Captions in the lower 15% of the frame, under the app buttons
Low-contrast captions with no outline
Subtitles that lag the voice by more than 4 frames