How to Turn a Long Video Into Shorts and Reels

To turn a long recording into shorts, describe the editorial result instead of hunting through the timeline. Video Trim watches the full source, finds the requested moments, and returns separate MP4 clips. You can ask for three 10–15 second moments, the exchange where two people discuss a topic, or one specific event with a target duration.

Scope note, September 2026: The workflow below uses the current Video Trim, Video Reframe and Captions pipelines. Video Trim now combines moment selection and trimming.

Step 1 — Put the source in your library

Upload the recording or import it from a URL first. Video Trim accepts a video asset from your library and does not need a separate transcript to find moments. Keep the original asset: every returned clip is a new reusable library item.

Step 2 — Describe the moments

Open Video Trim and write the request in the instructions field. Include the topic or event, how many clips you want, and a target duration when it matters:

• “Find the three most interesting moments, 10–15 seconds each.”
• “Find the part where the two speakers disagree about pricing.”
• “Cut the explanation of the launch plan to about 20 seconds.”

The agent reviews the complete source before choosing ranges. max_clips caps the number of returned videos, and max_clip_duration_sec prevents an individual result from becoming too long. The output is one library video per selected moment.

Step 3 — Review the clips

Treat the results as an editorial first pass. Check that each clip contains a complete thought, starts at a useful line, and does not repeat the same moment as another result. If the selection is too broad, make the topic more specific or reduce the maximum duration and run it again.

Step 4 — Reframe for the destination

A 16:9 recording cropped down the middle can put a speaker off-screen. Send the selected clips to Video Reframe with a 9:16 target for TikTok, Reels, or Shorts. It analyses keyframes and follows faces through the shot. Slides and screen recordings may be better with letterboxing than an aggressive crop.

Step 5 — Add captions

Use Captions when the clip needs burned-in text. If exact timing matters, run Transcription once on the source and reuse the SRT for the returned clips. Transcription is an optional supporting step, not a prerequisite for finding moments.

What makes the request work

• Name the editorial signal. “The strongest argument” is more useful than “make clips.”
• State the count and duration when you need a predictable batch.
• Describe one event when you want one clip; let the agent choose the clean boundaries.
• Use a maximum as a guardrail, not a promise. The agent can return fewer clips when the source does not contain enough distinct moments.
• Review every result. AI selection is a fast first pass, not a substitute for checking the final cut.

Try it

Start with a recording you already have and ask for three 10–15 second moments. If the results are useful, send those assets through Video Reframe and Captions. For a single precisely described exchange, ask Video Trim for that event directly instead of creating a separate highlights step.