Start with spoken footage and generate captions
Start with a spoken recording in the Cubix AI video editor. Settle the main cuts, then generate captions with Add Captions or the Generate captions control in Text → Subtitles. Listen to the recording while checking the generated words; timing and spelling still need review. These are candidate requests to adapt to your recording, not a tested recipe or a guarantee of a particular result. Replace every bracketed placeholder with a real preference, name, source quotation or visible step. Copy a request, open Cubix, add your footage and paste it into the editor conversation. Review the result before keeping it.
Four requests: restrained style, readable placement, names and a single revision
The four requests on this page cover initial styling, placement, a name correction and one style revision. State the preference you want and inspect the actual result; a request does not lock every other property or guarantee face clearance.
Asks for one style change. Compare the result with the previous edit and undo or adjust anything else that changed.
Use caption controls when a text or style change needs direct adjustment
Open Text → Subtitles, then Review every subtitle to inspect and edit caption text and timing. Fine-tune text exposes appearance controls; Words per phrase and Lines on screen control phrasing. Preview the actual recording while adjusting placement. You do not need to regenerate the whole transcript to correct a word.
Use direct caption controls for exact spelling, timing, phrasing or placement. If a conversational request does not give the change you need, use those controls and check the full result.
An illustrative transcript with a proper-name correction
In this fictional example, a caption reads “Dr. Sarah Almansor” while the correct name supplied for the recording is “Dr. Sara Al-Mansoor”. Open Text → Subtitles → Review every subtitle, find that line, edit the name and listen again. The names and transcript are invented to illustrate the correction, not an observed Cubix output.
This is a fictional spelling example. It does not report a successful prompt test or an actual transcription result.
Review spelling, timing, line breaks and faces before export
Perform a final manual check before rendering your clip. Verify that proper names match verified sources, subtitles remain legible against changing background colors, line breaks occur naturally at phrase boundaries, and captions never obscure the speaker's face.
- Verify names and technical terms against the recording or a supplied source.
- Preview the caption placement over the whole clip and check faces and important visuals.
- Inspect contrast, words per phrase and lines on screen at the viewing size.
- Listen for timing errors and read each line at normal playback speed.
✖ Relying on prompt requests for exact proper name spellings without verifying in text controls.
Why it fails: Automated speech engines may mishear uncommon names or technical jargon despite prompt hints.
✔ Better approach: Open the caption sidebar and manually edit proper names or technical terminology.
✖ Allowing multi-line captions to cover speaker faces on mobile vertical views.
Why it fails: Captions over a face can hide expressions or distract from the explanation.
✔ Better approach: State the desired placement, then inspect the actual frame and adjust the controls if needed.
✖ Over-styling captions with aggressive animations or rapid color flashing.
Why it fails: Excessive motion makes captions difficult to read and reduces professional video presentation.
✔ Better approach: Choose restrained, high-contrast typography with subtle highlight transitions.
Start this editing workflow in Cubix
Generate captions, choose a readable style and review text, timing and placement in the editor.
Editing workflows in Cubix
Frequently Asked Questions
Common questions around this editing workflow.