What The Levelator was used for
The Levelator was a free desktop application from The Conversations Network. You dropped a WAV or AIFF file onto it and it produced a new file with the levels between speakers and segments evened out. It had no settings, which was much of its appeal: it solved the specific problem of spoken-word recordings where one person is louder than another, without asking you to learn a compressor. Development ended years ago, and it is no longer maintained for current operating systems.
Pod Level as another option
Pod Level is built for the same narrow task: take one conversation recording, balance the speakers’ levels automatically, and give back a new file. There is nothing to configure. The things that can be said with confidence about how it differs come from how it is built:
- Nothing to install. It runs in current iPhone Safari, Chrome on iPhone, and desktop Chrome or Edge. Processing happens in the browser, and the audio never leaves your device.
- Compressed inputs. Alongside WAV, it accepts MP3, M4A/AAC, OGG or Opus, WebM, and FLAC files directly.
- Output format. It keeps the proposed format where the browser can encode it, or writes a smaller compressed file if you choose. The original is opened read-only and never modified.
- Current loudness targets. After balancing turns, the programme is normalized to −16 LUFS with true peaks limited to −1 dBTP, which is in line with common podcast delivery practice.
- Measurements and a preview. It reports input and output loudness, the gain range applied, and the number of turns adjusted, and lets you compare a few handoffs before and after.
- Free, with no signup. There is no account and no upload.
This page does not compare the two algorithms or claim one is better. The Levelator’s internals are its own; what Pod Level does is described below so you can judge whether it suits your material.
How Pod Level levels a conversation
A small voice activity model detects speech and splits it into turns. Each turn’s loudness is measured, and each turn is moved toward the recording’s typical speech level by up to 12 dB. Short or uncertain turns follow the gain of the speech around them, and gains ramp briefly at boundaries. Within a turn the gain is steady, so the voice is not compressed. The balanced result is then normalized and limited as described above.
Limits
Pod Level runs only in the browsers listed above; Android and Firefox are not supported. It processes one file at a time, does not reduce noise, and does not handle separate tracks per speaker. Large WAV output on iPhone is not supported, though compressed output is.
Using it
Open Pod Level, choose the recording, review the proposed output, pick a save location on desktop or start on iPhone, and wait for analysis and rendering. Progress and an estimate are shown, and you can cancel at any time. For more on the underlying problem, see balancing podcast speakers and fixing uneven podcast volume.