Make short clips from a video with AI
Choose passages that stand alone, then make reviewed horizontal or vertical clips from the approved main edit.
Make short clips from a video with AI when an extract needs a beginning the viewer can understand on its own. An interviewee might say “that’s what finally worked” after describing two attempts. The video-to-shorts workflow lets an agent map the words to programme time, review a longer or shorter candidate and preserve your choice until the final edit is available.
A portrait frame poses a separate editorial question. A speaker standing beside a printed diagram may fit neatly in the centre while the diagram vanishes outside the frame. Look at the beginning and end of the proposed crop. Where a useful composition cannot be verified, keep the horizontal version or request another framing pass.
What it does
- Compare extracts against the spoken words on the edited timeline.
- Check that a clip carries enough context to stand alone.
- Select more than one candidate without rendering every suggestion.
- Inspect a portrait frame for missing gestures, tools or captions.
- Export only confirmed ranges from the verified main delivery.
Install and use
- Use a Cut-as-Code project with the video-understand prerequisite, Python/Pillow and FFmpeg/ffprobe.
- Ask for passages that carry a complete idea and review the candidates beside the transcript.
- Once the main render is verified, check the selected frames and export the versions you approved.
FAQ
Source
Made by WhiteTowerAI. Original skill: video-to-shorts, published on GitHub under the MIT licence.
View on GitHub