Why Your Podcast Doesn't Sound the Way You Imagined — and What Auphonic (and Other Tools) Miss

You recorded an episode. You listened back. And you thought: this doesn't really sound like a proper podcast.

It's not your voice that's the problem. It's the room.

What nobody tells new podcasters

Most advice about podcast audio is about microphones. Buy a better mic. Hold it close. Use a pop filter.

All of that is true — but it solves the wrong problem.

What actually ruins podcast audio is acoustics: the hollow echo from a room with hard walls, the low hum from ventilation, the blurry "tail" that follows every word you say. You hear it clearly when you compare your recording to a professional production.

No microphone fixes that.

How the pros handle it

An experienced audio engineer — someone who has mixed live sound, broadcast, or TV — solves this with a chain of specialised tools. De-reverberation to remove room reflections. Noise reduction to remove background noise. Normalisation to hit the right level for Spotify and Apple Podcasts.

It takes time to learn. The tools cost money. And doing it right is essentially a full-time skill.

Most podcasters have neither the time, money, nor interest in becoming audio engineers. That's completely reasonable.

The automated alternative — and its limits

There are tools that automate parts of this. They're good at loudness normalisation: making sure your volume lands at the right level for different platforms.

But room reverb is a harder problem. It requires the model to understand the difference between your voice and its echo — and that's not the same as turning down a volume knob. Many automated solutions handle this with a general filter that works reasonably well on most files, but can create artefacts at the edges: voices that sound a little "processed", a little unnatural, a little robotic.

You don't always notice it immediately. But the listener does — even if they can't put words to what it is.

What PodMaster does differently

I've worked with audio for 20 years — live mixing, broadcast, TV production. That means I know what good audio actually sounds like, and what makes something feel natural or unnatural to a trained ear.

PodMaster is trained on recordings from real rooms: home offices, kitchens, bedrooms, conference rooms. Every recording has a "dirty" original and a version processed by professional tools as the reference. The model learns the difference — calibrated against human judgement of what sounds good, not just a mathematical formula.

It's two separate models in sequence: one for noise, one for room reverb. They solve different problems in different ways, just as an experienced producer would.

The result: you upload your raw recording and get back something that sounds like you recorded in a proper studio. Without needing to learn a single audio concept.

Where we are right now

The model is being trained during the summer of 2026. I won't make any promises about how good it is until I can show you an actual listening test — the same file, before and after, that you can judge for yourself.

That's planned for July. If you want to test it early: the waitlist is open.

Listening test update planned July 2026. Subscribe to the newsletter to get an email when it's published.