Get the words and picture working together.

Fix names, numbers, negatives and conditions before polishing the caption style.

Make the words arrive with the speech and remain readable; animation cannot repair a caption that is early, late or too brief.

Choose caption placement using the busiest frame and longest line, not just the opening shot.

Reduce unnecessary visual competition and phrase awkwardness; do not remove the words that qualify the speaker’s meaning.

Use a restrained style that stays legible across the recording; more motion is not automatically more useful.

Use verified names and specialist terms as a correction aid, then check that they were applied appropriately in context.

Use a verified name or a clear neutral role when speaker identity is necessary to understand the clip.

A translated caption needs a competent language review when accuracy matters; a fluent-looking sentence is not enough.

Ask for the exact deliverable: visible words inside a video, a timed-text file and a compositing overlay are not interchangeable.

Improve intelligibility, but do not use cleanup as a reason to guess unclear words or discard the original audio.