Video mode overview
Video mode is where Kitly does timeline work: cutting, dubbing, storyboarding, and AI-assisted paper edits, all over the same underlying sequence. Open it from the mode switcher and you land on the Sequences page, the home base for everything a project has cut, imported, or generated.
Not the same thing as Video tools
Video tools is the trim/crop/resize toolbar under a video preview in Sandbox and Asset Mode. It works on one clip at a time and never opens a timeline. Video mode is the full editor: multiple tracks, cuts, dubs, and export.
Sequences
A sequence is one timeline document. It can start life four different ways, and what it holds determines how it's labeled in the table:
| Kind | What it means |
|---|---|
| Edit | A blank timeline, built from New Edit. Drop assets on it and cut. |
| Redub | Created from a source video via New Redub. Carries the split stems, diarized speakers, and transcript that dubbing needs. |
| Assistant | A Redub sequence (or a container over several) that has an AI-generated cut library on it. |
| Storyboard | A sequence with shot panels on it, usually built up before there's a cut. |
A sequence's kind isn't a setting you pick and lock in. It's read off what the sequence actually contains, so a Redub sequence becomes an Assistant sequence the moment it grows a cut library, and reverts if you strip that library back off. Right-click a row for Change to Redub or Change to Assistant to do that by hand.
The table
Search narrows by name or kind, so typing "redub" filters down to redubs. Click a column header to sort; click it again to flip direction. Modified sorts newest first by default, everything else A to Z.
A single click on a row selects it. Double-click, or press Enter with it focused, opens it. This is deliberate: a lot of the checkboxes and the selection itself live on that same row, and having a single click open the sequence made it hard to select anything without accidentally opening it.
Batch actions
Tick a few rows and a bar appears above the table. Every button reports how many of your selection it will actually touch, so picking three Redubs and one Edit timeline and clicking Rerun Detection runs on the three Redubs and says so, rather than refusing the whole batch.
| Action | Applies to | What it does |
|---|---|---|
| Duplicate | Any sequence | A fresh copy, sharing all the generated audio and video underneath it. |
| Export ▾ Sequence FCP XML | Any sequence | Writes an interchange file per sequence into one folder you pick. |
| Export ▾ Sequence FCP XML + ZIP | Any sequence | The XML plus a copy of every referenced media file, zipped. A bare XML export still points at your machine's file paths, which is a headache the moment you hand it to someone else. |
| Export ▾ Sequence FCP XML + Copy To Directory | Any sequence | Same bundling, unzipped into a folder. |
| Rerun Detection | Finished Redub sequences | Re-transcribes every character clip using whatever speech engines Settings ▸ Local Inference currently resolves to. Timeline, splits, and translations are untouched. On cloud engines this is one call per clip, and it's billed. |
| Detect Scene Cuts | Finished Redub sequences | Re-runs local scene detection on the source video and re-cuts the video lane. Existing cuts near a detected boundary are kept. This is local ffmpeg analysis, not speech detection, and it costs nothing. |
| Run Split Clips | Finished Redub sequences | Sweeps every character-track clip and splits it wherever it finds a long-enough pause. Replaces trims and manual splits on those lanes. |
| Regenerate Version | Assistant sequences with a prompt in their history | Re-runs each sequence's most recent real prompt as a new draft. Existing drafts are kept, not overwritten. Each run is a model call and is billed. |
| Delete | Any sequence | Removes the sequence's timeline, cuts, and dubs. Source videos and any generated media stay in the project. |
Rerun Detection and Detect Scene Cuts are not the same thing
Both sound like "detection," but they're unrelated. Rerun Detection re-transcribes speech. Detect Scene Cuts finds video edit points. Pick the wrong one and you'll wait for a job that touches nothing you meant it to.
Anything that queues jobs or bills a model runs one sequence at a time, never in parallel, so a failure on one sequence never derails the rest and the progress bar means what it says.
Update Sequences
Word-level cutting needs a transcript with a timestamp on every word. Sequences imported before that shipped don't have one. Update Sequences, in the page header and again on an affected sequence's own toolbar, lists every Redub sequence with a status:
| Status | Meaning |
|---|---|
| Up to date | Already has timed words. |
| Can update | The import's own saved transcription is still on disk, so the words can be rebuilt from it with no re-transcription and no cost. |
| Needs re-import | The cached analysis is gone. Only a full re-import fixes this one. |
Running the update rewrites word timings only. It's deliberately not a re-import, which would rebuild the whole timeline and throw away every edit made since.
New sequences
The New menu covers every way to start one:
- New Edit: an empty timeline.
- New Redub: import a video for stem splitting, speaker detection, and transcription. Covered in full on Redub.
- New Assistant Cut: start from one or more already-analyzed videos, redubbed or not.
- Import from Project…: copy a sequence out of another project on disk, read-only against the source project.
Where to go next
- Cutting a timeline by hand: Editing
- Mixing audio on tracks: Mixer
- Planning shots before or alongside a cut: Storyboard
- Letting a model propose a cut from a transcript: Assistant Editor
- Dubbing, translation, and lip sync: Redub