How to Turn a Long Video into Shorts (by Hand and with AI)

Turning a long video into Shorts looks like an editing job. It is mostly a selection job. Cutting, cropping and captioning take minutes once you know which 30 seconds deserve to exist on their own; finding those 30 seconds is what eats the evening.

This guide covers the whole process, step by step, first by hand, then with AI, with the numbers from our own tests on real videos.

What makes a good Short

Before touching an editor, know what you are looking for. A clip that works in a vertical feed has five traits:

1. It makes sense to a stranger. No "as I said earlier", no reference to a slide nobody can see.
2. It hooks in the first two seconds. A question, a surprising claim, a strong reaction.
3. It carries one idea. One story, one tip, one answer. Two ideas make two clips.
4. It ends on a full sentence. Ending mid-word feels like an error, not a cliffhanger.
5. It is readable without sound. Most people scroll muted, so captions are not optional.

Method 1: by hand

Step 1: find the moments

The slow part. Three ways to do it faster than rewatching everything:

• Read instead of watching. Generate a transcript (most editors and YouTube Studio can) and skim it. You read 5 to 10 times faster than you watch.
• Mark during the recording. If you host, note the time when something lands. A two-word note at 23:40 saves twenty minutes of search later.
• Use your audience. On YouTube, the "most replayed" graph and timestamped comments point to the moments people already care about.

Step 2: cut on sentence boundaries

Start the clip on the first word of the hook, not on the breath before it. End right after the last word of the payoff. Remove the "so, um" at the start; keep the laugh at the end if there is one.

Step 3: reframe to 9:16

Crop the landscape image to a vertical 1080×1920 frame. Keep the speaker's eyes in the upper third, and leave room at the bottom and on the right, where TikTok, Reels and Shorts overlay their buttons and captions. If two people talk, either switch the crop to whoever speaks or use a split-screen with one person above the other.

A warning about sharpness: the vertical slice of a 1920×1080 landscape video is only 607 pixels wide. It will be scaled up to 1080 pixels on export. Filming in 4K gives you a 1215-pixel slice and visibly sharper Shorts.

Step 4: add captions

Burn captions into the image rather than relying on auto-captions, which viewers often don't turn on. Two to five words at a time, large, high contrast, away from the platform's buttons. Highlighting the word being spoken helps people follow.

Step 5: export and publish

Export in H.264 MP4, 1080×1920, at the frame rate of your source. YouTube Shorts and Instagram Reels accept up to 3 minutes; that is a ceiling, not a target. On YouTube, link each Short to the full video so viewers can go from the clip to the episode.

Method 2: with AI

AI tools automate step 1, which is the expensive one, and steps 2 to 4 with it. With Clipfox:

1. Upload the file or paste the link (YouTube, Twitch, Vimeo or most public links).
2. Choose the format (9:16 for Shorts), the number of clips (up to ten), and optionally a maximum length or an instruction like "only the parts about pricing".
3. Clipfox transcribes the speech, picks the strongest passages, cuts them on full sentences, reframes them on the speaker and burns in captions.
4. You review the clips, each with a title and a score, and download the ones you want.

What it gave on real videos

We measured it on two public-domain recordings (full results here):

The AI did not ignore the end of the videos: on the 29-minute show, three clips came from the last five minutes.

By hand or with AI?

• By hand when you post one carefully crafted clip a week, or when the best moment is purely visual (a reaction with no words, a sports action). You stay in full control of every frame.
• With AI when you want several clips per video every week, or when your videos are speech-heavy (podcasts, interviews, webinars, streams, lessons). You trade a little control for most of the time.

A good middle ground: let the AI propose ten clips, keep the five you like, and polish those by hand if needed.

Mistakes that kill Shorts

• Starting with context. "So, in the last episode we talked about…" loses the viewer. Start with the punchline or the question.
• Captions under the buttons. Check your frame on a phone before posting.
• One clip, one link, done. Shorts from the same video work as a series: publish them over several days.
• Promising quality the source can't give. A 720p webinar recording won't become a crisp 4K Short.

Next step: what to check before choosing an AI clip maker, or go straight to the use case that matches your videos: podcasts, YouTube videos, Twitch streams, webinars.

Turn your long video into Shorts with Clipfox: 30 free minutes every month with a free account.