Workflow5 min read

How to turn long videos into Shorts, Reels, and TikToks

One good long recording, a podcast episode, a webinar, a stream, a talking-head video, usually contains five or ten short clips that can carry your channel for a week. Most creators never extract them because the process feels like editing five videos from scratch. It is not. Clipping is one repeatable pipeline, and once you have run it twice it takes minutes per clip. Here is the whole workflow, start to finish.

Try Supercut freeNo upload · No sign-up to start · 100% private

Step 1: mark the moments while you watch

Play your long video once at normal or 1.5x speed with a notes file open. Every time something lands, write the timestamp and a two-word label. You are hunting for four kinds of moments: strong claims (anything that would make a stranger stop scrolling), emotional beats (laughter, surprise, frustration), concrete demonstrations (the before-and-after, the reveal), and clean self-contained explanations of one idea. Be ruthless about the "one idea" rule. A clip that covers two points holds nobody. Aim for moments that will cut to somewhere between 20 and 60 seconds, and collect more candidates than you plan to publish, because some will die in the edit when you discover the setup rambles or the payoff depends on context the clip cannot carry.

Step 2: cut each moment out

With timestamps in hand, the cuts are one prompt each. Drop the long video into Supercut and type the range: "trim from 12:40 to 13:15." The AI plans the cut, a deterministic engine runs it locally in your browser, and nothing uploads, which matters when the source is an unreleased episode or client material. Two trimming habits make clips noticeably better. Start the clip a beat after the thought begins, so the first words are already the hook rather than a wind-up like "so anyway, the thing I wanted to mention." And end it the moment the idea completes. Trailing seconds kill replays, and replays are what the algorithms reward.

Step 3: reframe to vertical 9:16

Long-form video is almost always 16:9, and every short-form feed is 9:16, so each clip needs reframing, not just posting. A straight crop ("crop to 9:16") works when the subject sits still in the center, like a locked-off talking head. When the subject moves, or the framing is wide, use smart reframe instead: "smart reframe to 9:16 and keep my face centered." Face tracking runs on your device and pans the vertical window to follow the subject, so a speaker who drifts across a wide stage shot stays in frame. Check the result for headroom and re-prompt if the crop sits wrong. If you are unsure which to use, smart reframe is the safer default.

Step 4: captions, because most viewers are on mute

A large share of short-form viewing happens with the sound off, so uncaptioned clips forfeit those viewers in the first second. Type "add captions, big bold style" and Supercut transcribes the audio with an on-device Whisper model, times each word to the speech, and burns the text into the frame. Because transcription is local, the audio never leaves your machine either. Skim the preview before export: proper nouns, brand names, and jargon are what any transcriber gets wrong, and they are exactly the words your audience will notice. Fix the text, then move on. Caption after reframing, not before, so the text is sized and placed for the vertical frame.

Step 5: export per platform and post natively

Export each clip as a 9:16 MP4 at 1080x1920, the size TikTok, Reels, and Shorts all want. Supercut has platform presets, so "export for TikTok" or "export for Shorts" gets the format right without memorizing numbers. Upload natively to each platform rather than cross-posting watermarked exports from one app to another, since platforms tend to suppress each other’s watermarks. The whole render happens in your browser, so there is no upload-render-download round trip per clip. Trim, reframe, caption, export, next clip. After the first one, each additional clip is the same four prompts with different timestamps.

What separates clips that work from clips that don’t

Format gets you to the starting line; these are the judgment calls that decide the race.

  • Hook inside two seconds: the first line on screen must create a question or a promise.
  • One idea per clip, stated fully. If it needs context from the long video, it is not a clip.
  • Cut the wind-up: start mid-energy, not at the beginning of the sentence.
  • End before it drags: a clip that feels slightly too short gets replayed, one that feels long gets swiped.
  • Readable captions: big, high-contrast, and out of the way of faces and platform UI.

Do it in Supercut

Related guides

Related use cases

Frequently asked questions

How long should clips from a long video be?

Twenty to sixty seconds covers most winners: long enough to deliver one idea with a hook and a payoff, short enough to hold attention and earn replays. Under 15 seconds usually cannot carry an idea, and past a minute retention decays fast unless the content is exceptional.

How many shorts can I get from one long video?

A one-hour conversation typically yields five to ten publishable clips. Mark every candidate moment on the first pass and expect to discard a third of them in the edit, when the setup or payoff turns out not to survive isolation.

Do I upload my long video to make the clips?

Not with Supercut. The full-length file loads into your browser and every trim, reframe, caption, and export runs locally through WebAssembly. Only your text prompts are sent so the AI can plan each edit, which matters when the source material is unreleased or client-owned.

Can I use the same clip on TikTok, Reels, and Shorts?

Yes. All three want vertical 9:16 at 1080x1920, so one export usually serves all of them. Export cleanly and upload natively to each platform instead of reposting a file that carries another platform’s watermark, which feeds tend to suppress.

What if the speaker moves around in the wide shot?

Use smart reframe instead of a fixed crop. On-device face tracking pans the 9:16 window to follow the subject, keeping a moving speaker centered without you keyframing anything by hand.