How to pull clips from a podcast episode without relistening to it

Read the episode instead of replaying it

The fastest way through audio you have already heard is to read it. A transcript lets you move at your own speed, search for a word, and see where an answer begins and ends without waiting for it to arrive. Thirty minutes of reading covers an hour of podcast, and you can stop on a line instead of scrubbing back four seconds at a time.

Work in sections of fifteen or twenty minutes rather than straight through. Attention runs out. An episode where the first twenty minutes got a careful look and the last forty got skimmed gives you a feed full of clips from the opening of every episode you publish, which is the least interesting part of most shows.

Reading costs you something, and it is worth naming. A transcript does not show the laugh, the pause, or the point where a guest's voice dropped. Those are often the whole reason a clip works. So read to find candidates, then listen to the candidates.

Write down the same four things for every candidate

A candidate is not a clip yet. It is a note. Keep the notes identical so you can compare them later without reopening the file.

• The first words, exactly as spoken
• A one-line summary of what the moment is
• A rough start and end timestamp
• Who is talking, when there is more than one voice

Use the summary as a filter

Two of those four do most of the work. The first words tell you whether the moment has an opening. The summary tells you whether it makes sense alone. If you cannot write the summary without the word 'basically', or without starting with 'earlier in the episode', the moment needs context a clip cannot give it.

Be willing to write down more candidates than you will use. Twenty notes for an hour is normal, and most of them will lose to something better. The notes are cheap; the cutting is not.

Compare the whole list before you cut anything

The temptation is to cut each good moment as you find it. Resist it. The third story in the episode is often a better version of the first, and you only see that with both notes side by side.

Sort the list roughly by how well the opening works, then strike the duplicates. Two clips making the same argument is one clip and one wasted slot. Keep the losers in a maybe pile, because next week you will want material and this is where it comes from.

This is also where you find out how many clips the episode really has. An hour with a guest who tells stories might hold a dozen. A looser episode might hold four. Forcing it to produce a dozen means posting eight things you do not believe in, and people can tell.

Move every start to the first word

A timestamp from a transcript is a clue, not a cut. Start a second late and you lose the subject of the sentence. Start a second early and the clip opens on a breath, a chair, or the tail of someone else's question.

Open the candidate, find the first word of the hook, and put the cut there. In an interview that is usually the guest's first word rather than the end of your question. When the question is short and makes the answer land, keep it, and then the cut starts on the first word of the question instead.

Do the same work at the end. A clip that stops before the payoff is worse than one that runs five seconds long.

Listen to the first three seconds of every export

This is the step people skip and the one that catches real mistakes. A timestamp can be right in a list and wrong in a file. Play the first three seconds of each finished clip. If the first word is there, and it is the word you expected, the clip is fine.

Then read the clip's own transcript against what you heard. A mismatch between the two usually means the cut landed on the wrong sentence, and that is far quicker to spot in text than by listening to everything again.

What the whole pass costs

By hand, an hour-long episode is an evening: reading the transcript, writing notes, comparing them, trimming the starts, exporting, and checking. The second episode goes faster. The tenth is still an evening.

EasyClip runs the same process on the file you upload. It works through the episode in sections, compares candidates across the whole thing, moves each start to the first word, cuts from your original file, and re-transcribes the opening of every export to check it against the line it was meant to start on. An hour-long episode uses 10 tokens for up to 10 clips, which fits inside the Starter pack (30 tokens for $10).

Audio-only shows work the same way. Upload an MP3, M4A, WAV, FLAC, or OGG file and you get audio clips back with transcripts, at the same 10 clips per hour. No video version is made from them yet, so there is no waveform and no burned-in captions. Good moments beyond the included clips are listed afterwards and can be added at 2 per token for 7 days, and slots the episode cannot fill come back as tokens.

Questions