1. The standard pass
Start with a clear camera-facing recording whose speech is understandable. This example asks for pacing, captions and restrained emphasis. Review the result against the source; it is not a tested generation or export recipe.
2. Pacing prompts
Decide which gaps can be shortened and which carry meaning. Start with a gentler request for a reflective explanation. Review quiet words, deliberate pauses and transitions; automatic silence removal does not guarantee removal of every filler or stutter.
- 'Shorten long waiting gaps but keep breathing room between sections.'
- 'Keep the pause after the question at 01:12.' — replace the timestamp with one you checked.
- 'Remove the false start before I say the main point.' — confirm the selected passage.
- 'Restore the cut that clipped the start of that word.' — use after reviewing the result.
3. Zoom and framing prompts
Tie emphasis to a real line or passage. Preview the crop throughout the clip; a request to keep the speaker centred is a goal to verify, especially when the subject moves or a second speaker appears.
- 'Add subtle punch-in zooms on my key claims, nothing aggressive.'
- 'Zoom in slightly when I mention the number, then back out.'
- 'At 5s snap to a tight shot, hold until 7s, then back to normal.' — a deliberate punctuation.
- 'I am sitting on the right and keep getting cut off — keep the frame on me.'
- 'Make this 9:16 and keep me centred even when I move.'
Ask for zooms on 'key claims' or 'when I mention X' rather than 'every ten seconds'. Emphasis tied to meaning reads as an edit; emphasis on a timer reads as a template.
4. Caption prompts
Generate captions from the recording, then review their text before styling them. Request the appearance you want and inspect phrase length, timing, contrast and placement on the intended screen.
- 'Add captions in a clean white style.' — restrained, professional.
- 'Add captions with a yellow keyword accent.' — emphasis without shouting.
- 'Add captions in a bold boxed style.' — short-form, high energy.
- 'Show two words at a time instead of a full line.' — phrasing, not styling.
- 'Make the captions a bit bigger.' — size only, look unchanged.
- 'In the captions, replace Jon with John.' — a text fix, no regeneration.
5. Uploaded cutaways and final review
Use supporting footage only when it clarifies the spoken point. Upload the file first and identify the intended placement. Before export, preview speech, text and framing together; check music levels manually if you add a track.
- 'Use my uploaded product close-up while I describe the product. Keep my voice underneath.'
- 'Return to the speaker at the end of that explanation.'
- 'Remove the zoom that crops the top of my head.'
- 'Move the captions away from the demonstrated object.'
✖ Using effects to hide an unclear opening
Why it fails: Visual emphasis does not supply missing context.
✔ Better approach: Choose a complete opening statement before styling it.
✖ Accepting captions without listening
Why it fails: Names and specialist terms may be transcribed incorrectly.
✔ Better approach: Check text and timing against the recording.
✖ Requesting unspecified B-roll
Why it fails: The editor needs the source and the intended placement.
✔ Better approach: Upload the file and identify where it should appear.
Start this editing workflow in Cubix
Try a focused request, then check the cuts, caption text and framing.
Editing workflows in Cubix
Frequently Asked Questions
Common questions around this editing workflow.