Caption proof: a ten-second sequence, checked before export

Steps
1. Mark the meaning-bearing action in each shot before placing text. In this example the hands fold a paper corner in seconds 0–4, press the crease in seconds 4–7, and reveal the result in seconds 7–10.
2. Transcribe speech and meaningful sounds. Correct names, punctuation and timing manually. Do not add a sound cue merely to fill silence; include it when a listener needs it to understand the scene.
3. Preview with the interface and playback controls visible on an actual phone. Move or shorten a caption that covers the crease. Do not crop the hands to make room for a fixed text box.
4. Watch once muted, then once with audio. A viewer should identify which corner to fold and when. Check the final export, since burned-in captions and a separate caption track can otherwise appear twice.

Worked record
00:00–00:04: ____________________
00:04–00:07: ____________________
00:07–00:10: ____________________
Muted review: ____________________
Export record: ____________________

00:00–00:04: “Fold the top corner inward.” Hands remain unobscured; caption above the working area.
00:04–00:07: “Press along the crease.” Keep the finger and crease visible.
00:07–00:10: “Open it to check the line.” Show the result before the next cut.
Muted review: Can a new viewer follow all three actions? Record the missed action and revise.
Export record: Filename, dimensions, caption-track choice, phone tested and reviewer decision.

The downloadable SRT uses this invented script so you can rehearse importing and correcting timings. It does not document a filmed test. If the caption must be shortened, retain the instruction’s meaning; if no placement leaves the action legible, choose a different shot. W3C guidance explains why relevant non-speech information can also belong in captions.

W3C: captions — https://www.w3.org/WAI/media/av/captions/