6 min read
How to start clipping in 2026: what to cut
A practical guide to clipping long video into vertical posts: which moments to cut, how many clips an episode really holds, and the mistakes that waste month one.
Clipping looks like an editing problem. It is not. Cutting one clip takes twenty minutes and most of that is mechanical. The expensive part is deciding which forty seconds of a two-hour conversation a stranger would watch to the end, and that decision is what stops nearly everybody in the first month.
Here is the process, without the part where somebody tells you that consistency is key.
Start from an archive, not from a plan
The most common failure is deciding to “start clipping” and then recording something new to clip. You already have the material. A podcast with twenty episodes contains roughly eighty to ninety publishable clips at the rate we measure on real projects, about four per hour of source. That is three months of daily posting sitting in a folder you have not opened since you exported it.
Working from an archive also removes the one variable that kills momentum: you cannot be blocked waiting for a recording. Pick the five episodes you remember as good, and start there. Your memory of which episodes were good is a better ranking signal than any score, including ours.
What a clip actually is
A clip that works is a complete thought that a stranger can enter cold. It has a beginning that establishes what is being talked about, a middle that goes somewhere, and an end that lands. It does not require the previous nine minutes.
That description rules out most of a conversation. Long-form talk is full of setup, agreement, tangents and shared context, all of it valuable in the room and useless in a feed. When people say a great episode “didn’t clip well”, this is usually why: the episode was good because of its arc, and an arc does not fit in forty seconds.
Things that clip well, in rough order of reliability:
- A specific claim, stated plainly, that a reasonable person might disagree with.
- A story with a turn in it, told in under a minute.
- A concrete number with its context attached.
- A correction: “everybody says X, and here is why that is backwards.”
- One sentence of genuine expertise delivered without hedging.
Things that almost never clip: introductions, mutual compliments, anything that begins with “so as I was saying”, and jokes that depend on a callback.
How many clips per episode, honestly
Across 548 real projects we average 4.3 clips per hour of source. A 65-minute episode yields four or five. A dense 90-minute interview might give seven. A rambling two-hour chat might give three.
Publishing that number is a choice. It would be easy to advertise “thirty clips per episode” and deliver thirty, because you can cut a two-hour recording into thirty forty-second chunks mechanically. Twenty-five of them will be filler, and filler posted daily trains an audience to scroll past you. The count is not the product. The selection is.
Plan your posting rhythm on the honest number: four clips per hour of recording, three platforms per clip, thirty days a month. A weekly two-hour show is about three posts a day. That is a real content operation, from one recording a week.
The three mistakes that eat the first month
Cutting on the video instead of on the transcript. Scrubbing a timeline looking for good moments is slow, exhausting, and biased towards whatever you happen to land on. Read the transcript. Good moments are visible as text, and you can read a two-hour transcript in fifteen minutes.
Writing the caption last. The first line of the caption and the first second of the clip are the same decision. If you cannot write a hook for a clip, the clip does not have one, and you should cut a different clip rather than write a better caption.
Batching to perfection. People spend three weeks preparing sixty clips and then post nothing while they keep refining. Post the first ten while you cut the next ten. The feedback from a real audience is worth more than another pass of colour grading, and you will discover within a week that half your assumptions about what works were wrong.
Vertical is not a crop
A 16:9 frame cropped to 9:16 down the middle puts a speaker’s ear in shot. Proper reframing follows whoever is talking and moves when the shot changes, not mid-sentence. If you are doing this by hand, set keyframes at the scene cuts and nowhere else: motion inside a sentence reads as a mistake even when the framing is technically correct. The rest of that mechanical pass, from the in point to the export spec, is laid out in how to clip a video.
Captions are not optional. A large share of viewing happens with sound off, and burned-in captions also keep a clip legible when a platform’s own captions are turned off or badly positioned. Keep them clear of the bottom third, where the interface lives.
When to stop doing it by hand
Manual clipping costs roughly three hours per episode once you are fast: reading, choosing, cutting, reframing, captioning, writing captions. At one episode a week that is a part-time job you did not apply for, and it is worth knowing what that hour is worth on the open market before you decide to keep doing it yourself.
The moment to automate is not when it becomes boring. It is when the backlog stops shrinking, when new episodes arrive faster than you clear old ones. Until then, doing it by hand teaches you what a good clip looks like in your own material, and that knowledge is what makes automated selection useful instead of noisy afterwards.
That is the whole method. Start with the archive you have, read rather than scrub, cut for complete thoughts, publish the honest number, and get the first ten out before you perfect the next ten.