Being based in Nashville is not incidental to what we built. This city has more working audio professionals per square mile than almost anywhere else in the country. Recording engineers, mastering engineers, live sound engineers, broadcast audio techs -- they are accessible in a way that is unusual if you did not grow up in a music industry city. When we were building Reverbwell and trying to understand what "good audio" actually means, we had the advantage of being able to ask people who do this every day for a living.
Several things they told us changed how we thought about the problem.
The room matters more than the microphone
Every podcast gear guide tells you to buy a better microphone. The audio engineers we talked to almost universally said the opposite: fix the room first, then worry about the mic.
This is basic acoustic knowledge in professional settings, but it does not make it into most podcaster advice. The reason: treating a recording space is harder to sell as a product. You cannot put a room on Amazon. You can put a microphone on Amazon.
What the engineers explained is that untreated room reflections -- the reverberant tail after a sound, the flutter echo between parallel walls -- are among the hardest audio artifacts to remove in post. Noise reduction can handle constant background sounds like HVAC or street traffic. It cannot cleanly remove room reflections without also removing parts of the voice that have similar spectral characteristics.
For Reverbwell, this meant we needed to train the noise model specifically on untreated home studio acoustics, not just on external noise sources. The recording environments our users actually work in are bedrooms with hardwood floors, closets with hanging clothes for dampening, spare rooms with minimal treatment. We recorded in those environments deliberately during the training data collection phase.
Level is a craft decision, not a technical one
Loudness normalization -- hitting a target LUFS level for podcast delivery -- sounds like a purely technical operation. You run the file through a tool, it hits the target, you ship it. Nashville mastering engineers corrected this assumption quickly.
The loudness target (-16 LUFS for mono, -19 LUFS for stereo on most platforms) is a delivery spec, not a creative spec. How you get there matters. A recording with heavy dynamic range compression applied early in the chain to hit the target sounds different from one where the gain structure was managed from the recording session forward. Both can hit the same number. Only one of them sounds like someone was actually paying attention.
This is relevant to how we built the level normalization in Reverbwell. Hitting the target LUFS is the baseline. Doing it without introducing pumping or over-squashing the dynamic range requires treating the loudness analysis and the processing as two separate questions. You measure first, then decide how to get there, preserving transients where the source warrants it.
What broadcast engineers know about remote audio that podcasters do not
Podcast remote recording -- two people on opposite sides of the country, each recording locally, tracks synced in post -- is a relatively new production standard. Broadcast engineering has been doing remote audio production for decades, and the approaches are different in ways worth understanding.
In broadcast, the standard solution for remote audio quality was to send a field engineer to the remote location with professional gear. That is not the podcast reality. What broadcast taught us instead was to think about the edit path before the recording session: what markers, what double-enders, what handclap slates, what level checks would make post-production tractable even with imperfect source audio.
That informed how we think about the Reverbwell track alignment algorithm. When two local recordings from a remote interview are uploaded as stems, the alignment problem is not just "find the common audio" -- it is "find the common audio accounting for the fact that one track may have been recorded at a different sample rate or with clock drift." Professional broadcast multitrack sessions assume synchronized clocks because they are running in the same facility. Remote podcast recordings do not, and handling that edge case required understanding it as a production problem, not just a signal processing problem.
The edit that sounds "mixed" versus the edit that just sounds "edited"
One conversation that stuck with us was with a mastering engineer who had done work for both music and spoken word. She drew a distinction between audio that sounds "mixed" -- where all the elements feel intentional, balanced, and of a piece -- and audio that just sounds "edited," where you can hear the seams, where the levels jump slightly between cuts, where the noise floor shifts.
Most podcast editing produces the second kind. Good podcast production produces the first kind. The difference is partly technical (level matching between cuts, matching noise floor in splice points) and partly a matter of paying attention to transitions, not just to the moments between them.
We built a pass into Reverbwell's processing pipeline specifically for level continuity across cut points. It is not a feature anyone asks for by name, but it is part of what makes the output feel like a finished edit rather than a cleaned-up rough cut.
What this means for podcasters who are not in Nashville
The specific geography matters less than the underlying insight: podcast audio is audio production, and audio production has deep engineering knowledge behind it that did not originate with podcasting. If you approach your recording and editing with some of that engineering thinking -- room before mic, gain structure before normalization, track continuity before export -- the output improves in ways that are audible to listeners even if they cannot name what changed.
We tried to bake some of that into Reverbwell so that the defaults are informed by production knowledge, not just by what is technically easy to implement.
