StreamLive Pro

How to turn a live stream into Shorts

Which moments of a live stream make good Shorts, manual vs AI time, reframing 16:9 to 9:16, Hebrew caption pitfalls, and official length limits.

Updated November 4, 2026 · 10 min read

An hour-long live class or interview usually holds a few moments that work on their own: a sharp answer, a short story, a clear piece of advice. Everything around them depends on context the viewer of a 40-second vertical clip will never have. Turning a live stream into Shorts is mostly the work of finding those moments and cutting them so they start at the right word.

This guide covers how to pick a moment, how long the manual and the AI routes take, how to reframe a wide shot into a vertical one, how to caption it (including Hebrew, which is easy to get wrong), the official length limits on YouTube, Instagram and TikTok, and the rights questions people skip until a guest complains.

What makes a moment worth clipping

Before any tool, there is a test. A good clip passes all four:

1. It stands alone. Someone who did not watch the live understands it. If the speaker says "as I said before" or "going back to the second point", the clip either starts earlier or is not a clip.
2. The point comes in the first two seconds. A question, a claim, a number, a surprising sentence. Not "so, um, let me explain". In a vertical feed the viewer decides almost immediately whether to stay, so the opening line carries the clip.
3. It ends on a complete thought. Cut after the answer lands, not in the middle of the next sentence and not 15 seconds later when the speaker drifts.
4. It is short enough to hold. For most talking-head material, 20 to 60 seconds is the range where one idea fits without padding. Longer is possible on every platform (limits below), but a 2-minute clip needs a 2-minute idea.

Where to look in a recording

• Answers to audience questions. The question is the hook, and the answer is usually tight because the speaker knows people are waiting.
• The moment the speaker changes pace. Lists ("three things you need"), a story with a punchline, a correction of a common mistake.
• Disagreement. In an interview, the moment two people see something differently is often good material for a clip.
• The first ten minutes are rarely it. Openings are greetings, sound checks and waiting for latecomers.

Manual workflow vs AI workflow

Both routes end in the same place: a vertical file with captions. The difference is who watches the hour.

The manual route

1. Watch or skim the recording and mark candidate moments (in and out times).
2. Cut each moment in an editor.
3. Change the canvas to 9:16 and reposition the picture so the speaker stays in frame, with keyframes where they move.
4. Transcribe, fix the transcript, style and place captions.
5. Balance the sound, remove dead air at the start.
6. Export, write a title and description, and upload to each network.

How long this takes depends entirely on you, so here is the arithmetic rather than a claimed average. Watching a 60-minute recording at normal speed takes 60 minutes; at 1.5x it takes 40. Suppose steps 2 to 6 take you 15 minutes per clip (time yourself on one clip and replace the number). Five clips is then 5 x 15 = 75 minutes of editing, plus 40 to 60 minutes of watching: roughly two hours per broadcast. Once a week, that is about eight hours a month.

The AI route

An AI clipping tool transcribes the recording, scores passages that look self-contained, cuts them, reframes them to 9:16 and burns in captions. Your part moves from making clips to reviewing them:

1. Upload the recording (or let the tool pick it up when the broadcast ends).
2. Wait while it processes. In our own system this is usually a few minutes per Short, and about twenty minutes for a set of five from a long video, as our help article says.
3. Watch each result once. Delete the weak ones, fix caption errors, adjust the crop where it is wrong.
4. Publish.

If review takes you 3 minutes per clip, five clips is 15 minutes of your time. The waiting time is real but you do not have to sit through it.

Which to choose

There is also a middle ground: let the AI make a first set, keep the good ones, and hand-cut the one moment it missed.

Reframing 16:9 into 9:16

A wide shot has room for two or three people side by side. A vertical frame is a narrow slice: about a third of the width of the original if you keep the full height (9/16 of the height is 0.5625 of it, and a 16:9 frame is 1.78 times as wide as it is tall, so the slice is 0.5625 / 1.78, about 32% of the width).

That number explains most reframing problems:

• One speaker, mostly still. Easy. Centre the slice on the face and leave it.
• One speaker who moves. The slice has to follow, smoothly. Speaker tracking (face detection plus smoothing) does this; by hand it is keyframes.
• Two people in one wide shot. A 32% slice cannot hold both. Options: follow whoever is talking (cut the slice between them on each change of speaker), stack them (top half and bottom half, each cropped to a square-ish frame), or keep the wide shot letterboxed in the middle with captions above and below. Following the speaker works best for conversation; stacking works when reactions matter.
• Slides or a screen share. Text on a slide becomes unreadable in a 32% slice. Either show the slide full-width in the middle of the vertical frame, or skip slide-heavy moments.

Plan for it while you stream. If you know clips will come out of the broadcast, keep the main speaker near the centre of the wide frame, and leave headroom. In a studio with 9:16 safe-frame guides, you can see during the live what the vertical crop will keep.

Captions

Some people watch short videos without sound, in public places or late at night, and captions also make a clip usable for deaf and hard-of-hearing viewers. We are not going to quote a percentage of muted viewing here, because we have not found a source we trust for it. Captions are worth it regardless.

What works in vertical clips:

• Short lines. Two to five words per caption line, not the two full lines a TV subtitle uses. The screen is narrow.
• Placement away from the platform UI. The bottom of a vertical video is covered by the caption text and buttons of the app, and the right edge by the like and share buttons. Put captions in the middle third or just below it.
• Highlighting the spoken word (karaoke style) helps viewers keep pace, and needs a transcript with timing per word, not per sentence.
• Burned in, not a separate file. On Shorts, Reels and TikTok, burned-in captions look the same everywhere. A caption file is right for long videos on YouTube (see our guide on translating captions).

Hebrew captions: what goes wrong

Hebrew runs right to left, but numbers, English words and some punctuation run left to right inside it. Caption tools built for English get this wrong in predictable ways:

• Punctuation on the wrong side. A full stop or question mark at the end of a Hebrew line shows up at the right, before the first word, instead of at the left. It happens when the line has no strong direction information at its end. The fix is an invisible right-to-left mark (Unicode RLM) at the end of the line, or a tool that adds it.
• Numbers and English words reorder a line. "We started in 2019 with OBS" in Hebrew can have the number and the English word swap places with the surrounding text when a line breaks in the wrong spot. Read every line that contains digits or Latin letters.
• Parentheses and quotes flip. An opening parenthesis can appear as a closing one.
• Fonts without Hebrew. A stylish Latin font with no Hebrew letters shows empty boxes. The tool must fall back to a Hebrew-capable font.
• Word highlighting in the wrong order. Karaoke animation that moves left to right across a Hebrew line highlights the words backwards.

If you caption Hebrew by hand, check the exported video, not only the editor preview: some players reorder text differently. Our help article on Hebrew captions for Shorts shows how this is handled in StreamLive Pro.

Platform specs: Shorts, Reels, TikTok

All three are vertical 9:16; 1080 x 1920 pixels is the usual export size. Length limits change often, so here is what each platform's own page said when we checked (4 November 2026):

The practical point: on YouTube a vertical video longer than 3 minutes is treated as a regular video, and on Instagram a reel longer than 3 minutes stops reaching people who do not already follow you. Keeping clips under a minute stays inside every limit above.

Posting cadence

There is no official number for how often to post. What we can say from the arithmetic: if you broadcast once a week and get five usable clips, you can post one clip a day on weekdays and never run out. That rhythm is easier to keep than a burst of fifteen clips on Sunday and nothing until next week.

• Post the strongest clip first, soon after the live, while the live itself is still being recommended.
• Same clip, all three networks is fine. Change the title and the first line of the description per network rather than the video.
• Remove platform watermarks before cross-posting. A TikTok logo on an Instagram reel tells viewers it was copied.

Rights: guests and music

• Guests. A guest agreed to appear in your live. That is not automatically consent to have a 30-second piece of their answer, out of context, travel on TikTok. Tell guests in advance that clips will be made, and remove a clip if a guest asks. For anyone under 18, get a parent's consent.
• Audience members. In a class or a synagogue service, people in the room did not choose to be on camera. Crop them out of clips or skip those moments.
• Music. Background music in a live, or the band at an event, does not become yours because it was in your broadcast. Platforms detect copyrighted music in Shorts and Reels and may mute or block the clip. Cut it or replace it with licensed music.
• Quotes from others. A clip of you reading someone else's long text is their content too.

What goes wrong

• The clip starts one sentence late. The hook was the question; the clip starts at the answer. Always check the first two seconds.
• The crop cuts off the speaker. Speaker tracking loses a face that turns away or leaves the frame. Watch every clip once before it goes out.
• A name is misspelled in the captions. Speech recognition does not know your guest's surname or a Hebrew term it has not seen. Names are a common caption error.
• Context lost. A sentence that was a joke in the live reads as a statement in a clip. If you would not stand behind the clip alone, do not post it.
• Fifteen clips at once. They compete with each other. Schedule them.
• Captions under the app buttons. Test on a phone, not on a monitor.

Where StreamLive Pro fits

StreamLive Pro makes vertical Shorts automatically when a broadcast ends: it picks moments, reframes to 9:16 following the speaker, burns in captions with Hebrew right to left handled properly, and adds a thumbnail. It does the same for any video you upload, with no live needed. The number of Shorts a month depends on the plan and is listed on the pricing page; the 7-day trial includes 10 Shorts with a small watermark.

The AI chooses moments, and it can choose badly or crop wrong. Each Short opens in the editor, where you can fix captions, move the crop, and export again, and you should watch every clip before publishing. More on the AI Shorts page, and the steps in Create Shorts from any video and Publish all Shorts at once.

Frequently asked questions

How long should a Short cut from a live stream be?

For a talking-head moment, 20 to 60 seconds usually fits one idea. YouTube Shorts can be up to 3 minutes and Instagram does not recommend reels over 3 minutes to new audiences, so under a minute is safe everywhere.

Can AI pick the best moments from my live stream?

It can find passages that look self-contained and make a good first set. It does not know your audience or the context, so expect to delete some clips and to fix captions, especially names.

How do I turn a 16:9 video into 9:16 without cutting people off?

For one speaker, follow the face with speaker tracking or keyframes. For two people, switch the crop to whoever is talking or stack them top and bottom. Slides need to stay full-width in the middle of the frame.

Why does the full stop appear on the wrong side of my Hebrew captions?

The line ends without direction information, so the player treats the punctuation as left-to-right. Adding an invisible right-to-left mark at the end of each line, or using a tool that does, fixes it.

Do I need my guests' permission to post clips?

Appearing in a live is not the same as agreeing to have a short piece travel on its own. Tell guests in advance that clips will be made and take a clip down if a guest asks.

Keep reading