How to Turn Long Videos Into Shorts (2026): A Practical Guide
Cutting a two-hour podcast into short clips is not hard. Cutting it into clips that people actually watch is a different job, and it's mostly about which thirty seconds you choose. Here's what matters, in order.
The part that decides everything: picking the moment
Most bad clips fail before a single edit is made, because the wrong thirty seconds were chosen. A clip works when it can be understood by someone who has no idea what came before it.
What that means in practice:
- Start at a complete thought, not mid-sentence. A clip that opens halfway through a word is a swipe.
- Look for a question, a number, or a claim. "Most people get this wrong", "It cost me four thousand", "Nobody talks about this" — these carry their own context.
- There has to be a payoff. A hook with no answer is the fastest way to train people to scroll past you.
- Avoid clips that need a name or a callback. If understanding it depends on knowing who "he" is, it won't travel.
The first two seconds do all the work
Short-form feeds are a swipe test. Most viewers who leave, leave almost immediately — and once they're gone the rest of your edit doesn't matter.
So: put the payoff at the start. No logo, no intro card, no "hey guys". If the strongest line happens forty seconds in, consider opening on it and letting the setup follow.
Captions are not optional
A large share of short-form is watched muted, especially the first few seconds. A clip without captions is a clip most people never hear.
What works: big, bold, high-contrast text in the lower third but above the interface controls. A few words at a time rather than dense paragraphs. Burned into the video, not uploaded as a subtitle file — platform captions are inconsistent and often off by default.
One practical trap: if your source video switches language, automatic transcription can switch with it mid-clip and produce captions in the wrong language. If you know the language, set it explicitly rather than leaving it on auto.
Framing: what to do with a 16:9 source
Vertical is 9:16 and your source almost certainly isn't. Your options are a centre crop, a crop that follows the speaker, or letterboxing with something filling the space.
For talking-head content a centre crop is usually fine and reads as intentional. Letterboxing with blurred bars wastes screen height that would otherwise carry your captions.
Titles, descriptions and hashtags
Write the title as the thought that makes someone stop, not as a description of the contents. "We were doing budgets completely wrong" beats "Podcast clip about budgeting".
Keep hashtags relevant and few. A dozen unrelated tags does not broaden reach — the platform works out what your video is about from watch behaviour, and confusing tags mostly confuse you when you're reviewing what performed.
The economics of the tools
This is the part that catches people who post at volume. Most AI clipping tools charge by credits — a monthly allowance of minutes or clips. That model is fine if you post occasionally. If you're clipping several long videos a week, you hit the ceiling and start paying per clip, right at the point where you're doing the most work.
It's worth knowing that the heavy parts — transcription and video encoding — run perfectly well on an ordinary PC. A modern machine transcribes and cuts without needing anyone's server. The credit model reflects how the tool is sold, not what the work costs.
Doing it on your own PC
That's why I built Clipwright: it takes a long video, transcribes it, finds the strongest moments, cuts them to 9:16, square or landscape, and burns in captions — entirely on your own computer. No credits, no monthly allowance, no cap.
Each video gets a folder with the clips, a thumbnail per clip, the full transcript, and a post-kit file. Connect a free AI model — Gemini, OpenRouter, Mistral and Cohere all have free tiers — and it also writes hooks, titles, descriptions and hashtags, and ranks the moments more intelligently. Without one, it still does all the clipping; you just write your own copy.
Your video never leaves your PC. If you connect a model, only the transcript text is sent to it — never the footage. Interface in English, Romanian, Italian, Spanish, German and French, and it's a one-time purchase. See the shop.
One thing worth saying plainly
Only clip content you have the right to use. Someone else's video is their copyright, and "the tool made it easy" has never been a defence. Clipping your own long-form, or content you're licensed or contracted to clip, is the version of this that doesn't end badly.
FAQ
How long should a short be? Long enough to land the payoff, short enough that nothing is padding. Twenty to fifty seconds suits most talking content.
How many clips from one video? Fewer good ones beat many weak ones. Three strong clips from an hour is a good outcome.
Do I need a powerful PC? No, but it helps. Transcription is the slow part.
Can I post the same clip to TikTok, Shorts and Reels? Yes, and most people do. Export vertical once and upload it to each.
More from the workshop: how this one-person studio works, or the rest of the apps.