How We Used AI in Podcast Mastering

The cover image for this post is by Growtika

When RJJ Software offered podcast production, mastering was part of the work, and we built our own tooling for it. The service has since closed, so this is written in the past tense.

What Mastering Involved

Mastering is the step that makes a finished episode sound consistent: levels balanced across segments, tone evened out, and the result sitting where a listener expects it whatever they play it on. Traditionally it is manual work, and it takes an engineer time. That time was the problem we set out to reduce.

What We Built

We built a set of in-house tools around machine learning models that we fine-tuned ourselves. They were not off-the-shelf products with our name on them, and they were not ordinary automation relabelled as Artificial Intelligence (AI).

The aim was to shorten the manual part of mastering, so that more of the production time went into the episode and less into the polish.

Whether it achieved that by any measurable amount, we cannot tell you. We recorded nothing at the time that would let us go back and check, and the tooling was retired along with the service, so there is nothing left to measure now.

A Note On This Post

This was published in January 2024 and edited in September 2026. The original version said the tooling offered unparalleled precision, that it surpassed human capabilities, and that it produced flawless audio. We measured none of it. Rather than leave those claims standing behind a closure note, we have taken them out.

ℹ️ This service is no longer offered

RJJ Software no longer provides podcast production, audio editing or mastering services. This post is kept as a record of how that work was done at the time. What that work was, and how it came to an end, is set out in The Podcasting Years.