Perpetual Chess Podcast: 500+ clips and 1.5M views, now three clips from every new episode

Perpetual Chess has been running for years. Ben Johnson sits down with a strong player every week and talks for an hour. By the time I started as his podcast clips editor, there were hundreds of episodes sitting in the archive, none of them clipped.

Over roughly three years I cut more than 500 clips out of that catalogue. They have done around 1.5M views in total. In 2026 the work moved from the archive to new episodes, in a new stacked format, three clips per episode.

A podcast is a harder clip source than it looks

Most chess content has a board to cut to. A podcast does not. You have two people talking, sometimes for forty minutes, and the entire value of any given moment is in what was said and who said it.

That removes every visual crutch. There is no brilliant move to land on, no reaction shot of a position turning. If the sentence is not interesting, no amount of editing makes the clip work.

So the job stops being an editing job and becomes a listening job. The edit is maybe twenty percent of the time. The rest is deciding what is worth cutting at all.

How I pick

Look for the claim, not the topic

A guest explaining how they prepare openings is a topic. A guest saying they stopped studying openings entirely and got stronger is a claim. Topics do not travel. Claims do, because a claim invites disagreement, and disagreement is the cheapest engagement there is.

The clip has to survive without the guest’s name

It is tempting to lean on who is speaking. Most of the audience scrolling past has no idea who any given grandmaster is. If the clip only works because you recognise the face, it will only work for people who already subscribe, which defeats the point.

Cut in at the sentence, not the breath

Podcast audio has a rhythm that makes it very easy to start a clip two seconds early. Those two seconds of throat-clearing are where the swipe happens. I start on the first stressed word of the actual sentence, every time.

What made it repeatable

Three years is long enough that taste alone stops scaling. I built a pass system: one listening pass at speed marking timestamps, one pass ranking those timestamps against the claim test, then batch editing only what survived.

Roughly one in six marked moments made it to a finished clip. That ratio has been stable for years, and knowing it is what makes the workload predictable instead of open-ended.

2026: three clips from every new episode

This year Ben moved me onto the new episodes as they come out. Each episode becomes three vertical clips, delivered as numbered files with a matching subtitle file into one shared folder. The current cycle runs from 1 to 12:

The stacked format

• Guest on top, Ben underneath. The side-by-side recording is split and stacked to fill a vertical frame. The letterbox and the panel split are measured again on every episode, because no two recordings are framed the same.
• A title lockup for the first four seconds: the guest’s title in red, the name in white, and a one-line hook under it.
• One to three words of caption at a time, sitting on the seam between the two faces, white for the guest and yellow when Ben speaks, so you always know who is talking with the sound off.
• Three to five real images per clip, placed on the exact words they illustrate and laid over the person who is not talking: the player being discussed, the tournament, the book. Every photo is free-licensed, and the sources are kept with the build.
• Silences tightened, never mid-word. Pauses are removed, but a cut never lands inside a word, and every join gets a short audio crossfade so it does not click.
• A freeze on the last frame with a “Watch full episode” card, so the closing captions are never hidden under it.

Every batch goes through a separate quality check before Ben sees it: clipped words at the joins, captions against the audio, image placement, loudness. The look is fixed in a build file, so clip 13 will look like clip 1.

Result

500+ clips, around 1.5M views in total, cut from a weekly show that was already being made anyway. No extra filming, no reshoots, no additional demand on the host. In 2026, three finished clips from each new episode, in a format his audience can recognise at a glance.

Thanks man, good job on the last clips.

Ben Johnson, Perpetual Chess Podcast

Tools

Premiere Pro and After Effects for the archive work. The 2026 clips are built by a script from one settings file per episode, with local transcription for word timing, which is what keeps three clips an episode consistent. Claude helps with first-pass transcript triage across long episodes. The judgement about which moments are worth cutting stays mine.

What a podcast clips editor should hand you

Clips that start on the sentence, captions you can read with the sound off, a format your audience recognises, and the subtitle files alongside so a typo is a two-second fix on your side. If your editor sends you a clip that starts on a breath, that is the first thing to fix.

More on the wider approach in chess content repurposing. If you have a podcast archive nobody has touched, or new episodes that need clips every week, book a call, or put it straight into the monthly Shorts-only package.