Match generated video to audio

Match these shots to the audio.

1. **Convert the narration to seconds** at a natural speaking pace, per sentence. Do this before touching shot lengths — audio length is fixed and picture is what adjusts.
2. Assign narration to shots. Where a shot is shorter than its narration, say whether to extend the shot or split the line.
3. **Lip sync** — mark every shot where a mouth is visible while someone speaks. *These are much harder than they look and usually better solved by not showing the mouth:* cut away, shoot from behind, or frame below the face. Say which applies.
4. Where silence should sit. A beat before a point lands does more than filling every second.
5. Sound beyond narration — ambience, effects, music — and whether the tool generates it or it is added in editing.
6. Where the cut should fall relative to the audio. Cutting slightly before the line finishes usually reads better than cutting exactly on it.

Then:
- Total runtime against the narration, and which shots to trim if it runs long.
- Shots whose length is driven by the picture rather than the audio, which can absorb adjustment.

Rules:
- *Estimate at a real speaking pace, not a reading pace.* Underestimating here is why generated video ends up rushed.
- Do not change my wording to fit a shot without saying so.