Editing Workflow

The transcript is the primary editing interface. This page covers how a transcript edit becomes a video cut, and how undo/redo works underneath.

Word-level timestamps

Every word in the transcript carries its own start/end timestamp, produced by Whisper at transcription time:

{
  "text": "Hello everyone",
  "start": 10.42,
  "end": 11.92,
  "words": [
    { "text": "Hello", "start": 10.42, "end": 10.91 },
    { "text": "everyone", "start": 10.95, "end": 11.92 }
  ]
}

This granularity is what makes transcript edits map accurately to video timestamps — a deletion doesn't need to round to the nearest sentence.

Transcript deletion → video cut

Deleted words → find start/end timestamps → create cut operation
             → update timeline → preview video

For example, deleting "um basically" from:

Hello everyone [DELETE] um basically [/DELETE] today we are learning AI.

where um spans 12.42–12.71 and basically spans 12.75–13.31, produces:

{ "type": "cut", "start": 12.42, "end": 13.31 }

The cut is applied to the timeline immediately — it's reflected as a gap on the timeline and skipped during playback, with a short crossfade across the splice point rather than a hard cut, so the audio doesn't pop.

Supported transcript actions

• Select and delete arbitrary text
• Delete a whole sentence
• Delete a whole transcript segment
• Jump from transcript position → video timestamp
• Jump from video playhead → transcript position

Undo, redo, and restoring a specific cut

Every edit is represented as an operation, applied in sequence. Undo reverts the latest operation; redo re-applies it. Beyond plain undo, you can also restore one specific previously-cut word, segment, or sentence directly — without having to undo everything after it.

The original uploaded video is never mutated for any of this — cuts are timeline state (which source ranges map to which timeline ranges), not edits to the underlying file. See Architecture for the Track/Clip model this is built on.

Timeline view

Alongside the transcript, the timeline shows zoomable frame thumbnails and a live-updating waveform. Both get denser (more thumbnails, more waveform detail) as you zoom in, rather than just stretching existing pixels — useful for finding a precise cut point by ear or by eye.