Video assetization editor
Opening a video assetization project opens this editor. It is where people confirm the lines and speakers the machine separated, and line the timing up on the waveform.
The screen#
| Area | Description |
|---|---|
| Top | The source player |
| Center | The waveform, with Dialogue and Captions on their own tracks |
| Bottom | The line list — timecode, source text, and speaker on one row |
| Right | Evidence, Segments, On-screen text, Issues |
Confirming speakers#
Every row shows where its speaker stands.
| State | Meaning |
|---|---|
| Confirmed by a person | Someone looked and decided |
| Machine guess | Diarization output. Not review yet |
| Same speaker as above | Carried over from the previous line |
| No speaker assigned | Nobody stands here yet |
When choosing a speaker, look In this title or Across the workspace, and narrow by name. When one person stops and another starts mid-block, use Different speaker from here.
Evidence panel#
Select a line and the Evidence for its speaker appears — voiceprint, sample count, similarity range, face keyframes, appearances elsewhere. Evidence is machine-derived, so people do not edit it.
Working the lines#
| Action | Description |
|---|---|
| Merge into the previous line | Rejoin a line the machine split wrongly |
| Split the line at the caret | Separate two people's words in one row. The second half has no speaker yet |
| Edit the source text | Double-click or press Enter |
Timing on the waveform#
Drag a block to move it, grab an edge to resize, double-click to edit the text.
Timecodes read in Frames or Milliseconds. Switching also changes how far ⌘← and ⌘→ nudge.
The waveform is not always drawn.
- The source is DRM-protected, so it cannot be read from audio
- The episode is too long to read from audio; a pipeline waveform is needed
- Audio could not be read
Segments#
A segment boundary is where the place changes. Out of the scenes the machine found at cut points, it draws dividers only where the place changes. People then split, merge, and name the place and the action line.
On-screen text#
Text that appears on screen lives on the Captions track. Pick its kind — Name caption, Corner title, Crew comment, Sign.
A Name caption can be pushed onto the segment's speaker with Use as the speaker. Separating name captions cuts speaker work.
Filter and issues#
Combine the speaker-status and sentence-status axes to isolate lines that are done but still have no speaker.
Issues are raised as Wrong transcript, Wrong speaker, or Segment boundary. Pick the sentence first.
Shortcuts#
| Shortcut | Action |
|---|---|
| ↑ ↓ | Previous line · next line |
| Enter | Edit the source text |
| ⌘← ⌘→ | Nudge the start or end one step |
| Esc | Close the open panel |
After completing#
Once the assetization project is complete the screen becomes Completed · read-only. Its results are already the source of truth, so even the waveform can no longer be edited.
Account and contact details in screenshots are anonymized. Menus may vary by role and release.