Why your Substack podcast sounds worse than it reads

The thing listeners hear as amateur is almost never the microphone. It is the room: a hard, empty space bouncing your voice back at the microphone a few milliseconds late, which the ear reads as distance.

Soft surfaces fix it and they are free. Curtains, a rug, a sofa, an open wardrobe of clothes behind you. Record thirty seconds of silence in your space and play it back loud. Whatever you hear is on every episode.

Buying a better microphone before fixing the room makes the problem more audible, not less.

You lean in when you are interested and back when you are reading. Over forty minutes that produces a recording where some passages are twice as loud as others, and a listener in a car reaches for the volume control repeatedly. They do not consciously notice, and they do not finish the episode.

This is entirely fixable after the fact with levelling and loudness normalisation, and it is the single change that most makes a recording sound produced.

This one is not technical and it is the hardest to hear in yourself. Writers read their own prose at the punctuation rather than at the meaning: a short pause at every comma, a longer one at every full stop, evenly, like a metronome.

Real speech does not do that. Some commas want nothing. Some sentences want two full seconds afterwards because the idea needs to land. Mark up your draft before recording with the three or four places you actually want to slow down, and ignore the rest of the punctuation.

Sentences written for the eye often cannot be said at all. If you stumble on the same clause twice, the sentence is the problem, not your delivery.

Most writer podcasts open with housekeeping: who they are, what the newsletter is, a thank you to subscribers, a note about last week. A stranger in a podcast app gives you about ten seconds and all of it went on admin.

Start with the most interesting sentence in the piece and do the introductions afterwards, once they have a reason to care who you are. This costs nothing and is the highest-leverage edit available.

Three seconds of nothing before you start, and a long tail after you finish where you reached for the stop button. Both read as unfinished, and both take ten seconds to trim.

Trim to roughly half a second at the front and end cleanly on your last word. Do not fade out. A fade on a spoken piece sounds like a mistake being covered.