Most people watch with the sound off for the first few seconds. The captions are the short until they tap.
Three rules
Three words at a time, so they can be read at a glance. The middle third of the screen, above the app’s buttons and below the notch. Never over a face or a product.
We set captions in your fonts once, then check every line by ear. A machine gets the words right most of the time; a person gets the names right.