Transcript editing
How text edits map to frame operations.
The transcript is aligned to the video at word level. Editing text is editing video.
Operations
| You do | The pipeline does |
|---|---|
| Delete a sentence | Removes the frame range, snapping cut points to the nearest breath |
| Reorder a paragraph | Moves the corresponding segment, re-timing everything after it |
| Correct a word | Re-synthesises that span in the current voice and splices it in |
| Insert a new sentence | Generates the audio and holds the last frame or inserts b-roll |
| Mark a span as emphasis | Adjusts delivery and optionally adds an on-screen callout |
Cut-point snapping
Cuts are never placed exactly at a word boundary — that produces audible clicks. The pipeline finds the nearest low-energy point within 180ms and crossfades room tone across the join.
Keyboard shortcuts
| Shortcut | Action |
|---|---|
| ⌘/Ctrl + K | Command palette |
| ⌘/Ctrl + D | Delete selection and its frames |
| ⌘/Ctrl + Z | Undo (restores frames exactly) |
| Space | Play / pause from the caret |
| ⌘/Ctrl + J | Insert a zoom keyframe at the caret |
| ⌘/Ctrl + Enter | Ask Riff about the selection |