Voiceover direction that stops the robot read

Prepare this script for AI voiceover: {{paste script}}.

Do three things:
1. Rewrite for the ear: split any sentence over 20 words, replace words that TTS mangles (numerals, symbols, acronyms, units) with spoken forms, and remove parentheticals that break the line's rhythm.
2. Mark the performance inline: [pause 0.5s] where a beat is needed, CAPS on the single stressed word per sentence, and note where the pace should drop for the important claim.
3. Give the voice direction as a short brief: age range, energy level, warmth, pace in words per minute, and two reference descriptions ('like someone explaining something they actually care about to one person, not an audience').

Then list: the words most likely to be mispronounced with the phonetic spelling to use, and the lines that will still sound synthetic if the delivery is flat.