# How I cut a podcast episode into clips (the boring, honest version)

_September 20, 2026 · Clips Done · https://clipsdone.com/blog/how-to-turn-a-podcast-episode-into-short-clips_

I cut clips for podcasters every day, and the same five things decide whether a clip gets watched or scrolled past. None of them are clever. Most of them are just patience. Here's the whole process, so you can do it yourself or so you know what to check when someone does it for you.

## Read the transcript, don't watch the episode

Watching a 45-minute episode to find clips takes 45 minutes. Reading the transcript takes ten, and you notice things you'd never hear: a sentence that's quotable on its own, a story that starts and ends in under a minute. I mark anything with a number in it, anything where the guest disagrees with the host, and anything that answers a question people actually type into Google.

Last week a marketing podcast had a guest tell a story about two air-conditioning companies who both lost a job because nobody picked up the phone. That's a clip. Three minutes later she explained her general philosophy of customer service. That's not a clip, even though it was the "important" bit. Nobody stops scrolling for a philosophy.

## Cut on sentences, and give it air

If you take one thing from this page: start on the first word of a sentence and end on the last word of one. Then leave a quarter of a second of silence on each side. Auto-clippers cut where a timer tells them, which is why half the clips on your feed open with "…and so basically" and end mid-word.

Thirty to sixty seconds is the range that works. Under twenty and it's a fragment. Over ninety and people leave before the payoff, which is a shame because the payoff is why you picked it.

## Point the frame at the person talking

Vertical video shows about a third of a normal landscape frame. So the question for every second of the clip is: which third? Two mistakes happen constantly.

The first is **drift**. Face-tracking software slides the frame every time the speaker leans. It feels like watching from a boat. The fix is dull: lock the crop for each camera shot, and only move it when the camera actually cuts.

The second is **the wrong person**. In a wide two-shot the crop has to jump to the guest when the guest speaks, and jump back when the host replies. If it lags by three seconds, the viewer watches someone nod for three seconds. I check this by scrubbing with the sound off. If I can tell who's talking from the frame alone, it's right.

For audio-only shows, use the cover art and a moving waveform, and let the captions do the work. A static photo behind text looks like a slideshow, because it is one.

## Captions: read them against the audio, every time

Something like seven in ten people watch short video on mute at first, so the captions are the clip. Automatic transcription is good now, and it still gets names, product names and numbers wrong. I've seen "Kubernetes" become "cooper nettys" and a co-host called Dargie come out as "Darcy" in every single caption, which he noticed before the host did. Read every caption once with the audio playing. It takes two minutes per clip.

Two to four words at a time, big, and highlight the word being spoken. The highlight isn't decoration, it keeps the eye on the text so people don't drift off.

## Title it with what they said

The on-screen title is the thing people decide on. Use the speaker's words, shortened. "Everything that moves will be robotic" stops people. "Jensen Huang on the future of robotics" doesn't, because it's a description, not a claim. Put it in the top third, above the head, so it shows in the feed preview before anyone taps.

## Write the post once

Each clip needs a title, one line of post text, two sentences of description with the episode name, and a few hashtags. Write them together while the clip is fresh, then paste the same set into Instagram, TikTok, Shorts and LinkedIn. One episode gives you eight to twelve clips. At three posts a week, that's a month of content from one recording.

## What it actually costs in time

Done properly, a 45-minute episode is about three hours of work: an hour reading and picking, an hour cutting and framing, an hour on captions, titles and post text. That's the honest number. It's why most hosts either stop after a month or post whatever the tool spat out.

---
Clips Done cuts podcast episodes into 8 to 12 finished vertical clips, checked by a person, within 24 hours. $99 an episode or $350 a month with posting. First episode free: https://clipsdone.com