Caption creation starts from a video layer whose asset reports an audio track. The selected video interval determines which transcript cues are used. Create captions at 1× for the most predictable timing; captions created from speed-adjusted video require a complete synchronization review.
Create captions from narration
Select the source layer
Choose an audio-capable video in the timeline. Use 1× while creating captions; for a sped-up or slowed-down clip, normalize the rate or verify the full caption range carefully after creation.
Open Captions
Expand Captions in the source inspector. Wait for available subtitle metadata to load.
Apply the available source language
Choose the auto-detected source entry only after it is shown as available. The editor requests cues for the selected clip’s in and out range.
Wait for the caption layer
Successful cues are clamped to the source interval, converted into timed word tokens, stored as a caption asset, and inserted directly above the source track.
Verify the result
Seek into several phrases and compare audio, displayed words, timeline interval, and token times.
Default layer geometry
The initial caption box is centered horizontally, placed below composition center, limited to a practical width, and sized for two lines. Defaults include a bold TikTok Sans treatment, white text, bright highlight, black stroke, centered alignment, and a two-second page duration. Treat them as a readable starting point—not a final design.
Range behavior
Only cues intersecting the selected source interval are kept. Cue starts and ends are clamped to the range, empty or zero-duration cues are discarded, and sentence-like cues are split into word-level time segments for page-based rendering.
Caption-source link
Creation links the new caption layer to its source video in both directions. That relationship keeps ordinary moves and nudges together. Trimming the video can propagate to its captions; trimming the caption alone does not change the video. The timeline link control can unlink the pair.
Deleting either member of an intact source-video/caption pair deletes the linked member too. Unlink first when you need to remove or restructure only one. Also unlink before splitting a linked video/caption pair: a split does not rebuild the source relationship across the resulting pieces. Keep the pieces independently timed, or recreate captions from the intended source video when you need a linked caption layer.
Language options that are pending, processing, or not yet successfully applied are not verified caption data. Do not time the edit around them until a caption layer is created and audibly checked. Treat captions from any non-1× source as unverified even before later speed changes.
