Last month I watched an editor hand me two cuts of the same 58 minute interview. Same footage, same audio, same speakers, same software. One ran 44 minutes and felt like a conversation you wanted to finish. The other ran 41 minutes, was technically cleaner, had fewer stumbles, tighter gaps, and it was unlistenable. Not bad. Just dead.
Nothing on a checklist explains that gap. Both cuts passed every rule you could write down. Levels were matched. Filler words were gone. Sync was frame accurate. Cameras cut on the right speaker. If you graded them on a rubric, the shorter one wins.
That is the whole problem. Podcast editing is not a technical problem that occasionally requires taste. It is a taste problem that happens to require technique. People get this backwards constantly, and it is why so many shows sound polished and feel like nothing.
The checklist is the floor, not the work
I am not anti process. We run one. Every episode gets watched end to end by a human, and I mean actually watched, not scrubbed. Nothing gets auto-cut by silence detection, because a silence detector cannot tell the difference between a host losing their place and a guest deciding whether to say the true version of the answer. Those two silences look identical on a waveform. They are opposites.
So yes, there is a floor. Sync, levels, noise, breath, cameras, sequence. If you cannot hit those reliably you have no business talking about taste, in the same way a chef with a dirty kitchen does not get to talk about seasoning. I have fixed enough genuinely broken audio to be sure of this: some problems are not taste, they are damage, and you fix damage first.
But the floor is where most editing advice stops. Remove the ums. Trim the pauses. Cut anything that is not on topic. Keep it tight. That advice produces a specific kind of episode, and I can hear it within ninety seconds. It sounds like someone reading a transcript out loud at 1.1x speed.
What taste actually is in an edit
Taste is not a vibe. It is a set of decisions you can name, you just cannot pre-write them.
Here is what I mean concretely. In one episode this year, a guest gave an answer, stopped, and said "actually, no, let me say that properly." Then gave a better version. The rule says cut the false start and keep the good take. We kept both. The false start is the moment the listener realises this person is thinking in front of them and not reciting. That eleven seconds is the most valuable eleven seconds in the episode and every rule I know would have deleted it.
Another one. A host laughed over their guest's punchline and buried it. Technically that is a fix: duck the laugh, recover the line. We left it loud. A show where nobody ever steps on anybody is a show nobody is enjoying.
Taste is knowing which of these to do:
- Keep a stumble because it proves the person is real
- Cut a perfectly fluent sentence because it says nothing
- Hold on a reaction for two extra seconds instead of cutting to the speaker
- Let a pause run long because the answer that follows deserves the runway
- Kill your own favourite joke because it breaks the momentum of the segment
None of those are in a manual. All of them are choices, and each one has a right answer for that specific show that is wrong for a different show.
The tightest possible cut and the best possible cut are almost never the same cut. If your editor only knows how to make it shorter, you have a trimmer, not an editor.
The tell is always the filler words
If you want a fast read on whether an edit had taste applied, listen to how the ums were handled.
Bad filler removal has a signature. The voice steps. You hear a tiny gear change where the breath got amputated and two half breaths got glued together. It happens most on the words either side of the cut, and it happens because someone was removing filler as a search and replace operation rather than a listening one. I wrote a whole piece on cutting filler words without making the audio sound robotic because it is genuinely the single most common thing I hear in shows that are otherwise fine.
The taste version is simpler than people expect. You do not remove every um. You remove the ones that were doing nothing. The um before a hard question, the one where the guest is buying two seconds to decide how honest to be, that one stays. It is punctuation. It tells the listener something is coming.
Same with punch ins. A punch in that lands feels like the show noticed something. A punch in that does not feel motivated just reads as an editor trying to look busy, which is a separate craft problem I got into in punch ins that do not look cheap. The camera cut is not decoration. It is emphasis, and emphasis you use on everything is emphasis you use on nothing.
Taste is per show, not universal
The thing that makes this hard to systematise is that taste is not one taste. It is a read on what a particular show is for.
A two host banter show and a founder interview show want opposite things from the same raw moment. The banter show wants overlap, speed, interruption, mess left in. The interview show wants space, patience, and the discipline to let an answer land without cutting away to a reaction shot. Apply the banter edit to the interview and you get something that feels anxious. Apply the interview edit to banter and you get something that feels like a hostage video.
So the first thing our editors do on a new show is not open the timeline. It is watch two or three of your existing episodes, if you have them, and figure out what the show already is. Not what it should be. What it is. Then we edit toward that, harder.
This is also why we do not hand a new client a template and hope. It is baked into how we work an episode, and it is why the first cut lands in 48 hours rather than four. Two of those days exist so a person can form an opinion about your show before touching it.
Where taste matters most: the first ninety seconds and the clips
If taste only got applied in two places, I would spend it here.
The first ninety seconds. Almost every show I get sent opens with roughly a minute of throat clearing. Welcome back, before we start, quick housekeeping, and by the way. Meanwhile the actual reason to listen is sitting at minute 34. Moving it is not a technical act. It is a judgement call about which sentence in a 60 minute file is the one that earns the next 59. That is the entire skill in podcast trailer editing, and the skill transfers directly to the top of the full episode.
The clips. Vertical clips are where taste is most visibly absent in the industry. The default approach is to find a moment where someone says a punchy sentence, top and tail it, add captions, ship. That produces clips that are technically correct and emotionally flat, because the punchy sentence usually needs the setup line before it to mean anything. A clip is not an excerpt. It is a tiny episode with its own beginning.
We treat Instagram reels and podcast clips as their own edit, not as a cropping job, and it is the difference between a clip that gets watched and a clip that gets scrolled. Same principle applies on YouTube, where the packaging carries more weight than most hosts want to admit. Our sister studio ClickTheory does the thumbnails for exactly this reason: a great edit behind a bad thumbnail is a great edit nobody clicked.
Can taste be bought, taught, or automated
Automated, no. Not the interesting part.
Machines are already better than humans at the floor. Transcription, levelling, noise reduction, speaker detection, rough assembly, all of it is faster and increasingly more accurate than a person with a mouse. I use tools. I am not romantic about doing things by hand for its own sake.
But every automated cut I have tested makes the same category of error. It optimises for tightness because tightness is measurable and interest is not. It removes the pause, keeps the fluent line, cuts on the word boundary, and produces something that reads perfectly and feels like a press release. The machine has no opinion about your show. Taste is an opinion.
Taught, partly. You can teach someone the vocabulary of decisions fast. What takes time is the calibration, the part where you know that this particular pause on this particular show should stay. That comes from watching a lot of episodes all the way through and being wrong in front of someone who tells you.
Bought, yes, and this is the part I will be blunt about. Taste is what you are actually paying an editor for. The technical work has a floor price because it is commoditised. The judgement does not. When someone asks me why our pricing sits where it does, at $199 an episode for a single camera 1080p cut and $299 for the multi camera 4K version, the honest answer is that the difference between those numbers and a cheaper number is not more effects. It is a human who watched the whole thing and formed a view.
How to tell if your editor has taste
You cannot audit this from a features list, so here is what I would actually check.
- Give them one episode and ask what they would cut and why. Not how much. Why. A good editor names moments. A weak one names minutes.
- Look at where they cut, not what they cut. Cuts inside a thought are cheap. Cuts between thoughts are edited.
- Listen for whether the show still sounds like the host. If the edit made everyone sound the same, the edit had a personality and your host did not.
- Ask what they left in. This is the best question. Anyone can list removals. Only someone paying attention can defend a keep.
- Check the clips against the episode. If the clip is just the loudest thirty seconds, nobody made a decision.
The market is full of shows now, and it is going to stay that way, which I got into in why the podcast boom is not over. More shows means the technical floor stops being a differentiator, because everybody clears it. What is left is whether the edit has a point of view.
What I would do about it this week
If you are running a show and this is landing uncomfortably, the fix is not a new plugin.
Sit down with your last three episodes and write one sentence for each: what was this episode actually about. Not the topic. The thing a listener would tell a friend. Then go find where that moment sits in the timeline. If it is at minute 30 in all three, you do not have an audio problem, you have an ordering problem, and no amount of noise reduction touches it.
Then pick your worst performing episode and re-cut just the first two minutes. Nothing else. See what happens.
That exercise is basically the same one I run when someone asks me how I would start a podcast today. Get the judgement right and the rest is production. Get the judgement wrong and production just makes the wrong thing shinier.
I am in Dubai and we work with hosts everywhere, and the shows that grow are almost never the ones with the best gear. They are the ones where somebody decided what the show was and edited like they meant it. If you want a second opinion on your last episode, send it over. I will tell you what I would keep.