Shorts from Long Video

In short

A long lesson or broadcast contains several strong standalone moments, but manual searching, deduplication, cropping, and subtitling take more time than editorial selection.

Outcome

The pipeline builds a word-by-word transcript from an authorized video, ranks moments by explicit criteria, removes overlaps, and renders vertical drafts to a closed review folder.

How the automation runs

Trigger

Owner added video with reprocessing rights and launched candidate creation

Automation steps

1. Checks rights, duration, sound, and source composition
2. Builds word-by-word transcript and scene map
3. Ranks standalone moments by two explicit layers of criteria
4. Combines overlapping candidates and saves timecodes
5. Renders vertical drafts with verifiable subtitles to a review tray

Human check

The editor verifies rights, context, meaning, cropping, and subtitles, and manually selects for publication; the pipeline does not publish, clone identity, or change the speaker's meaning.

Outcome

The pipeline builds a word-by-word transcript from an authorized video, ranks moments by explicit criteria, removes overlaps, and renders vertical drafts to a closed review folder.

Using it

When to use it

You have your own archive of long videos and need a regular stream of short candidates without automatic publication and loss of context.

How to verify

Candidates have different timecodes and explainable scores, overlaps are collapsed, speech and subtitles match, context is self-contained, and no draft is published automatically.

Tools

Recipe details