Podcast Loudness and Noise Floor Conflict
A two-host podcast episode runs 48 minutes. Host A records on a dynamic microphone through an interface at a healthy level, with a room noise floor around -55 dBFS. Host B records on a USB microphone in a different room with a laptop fan audible and a noise floor around -38 dBFS. The hosts speak at different average levels, and Host B sometimes leans away from the mic. The editor must deliver a stereo file at a platform-recommended loudness of -16 LUFS integrated with a true peak ceiling of -1 dBTP. During a test mix, the editor normalizes the whole file to -16 LUFS and finds Host B's room noise becomes noticeably audible between phrases, and Host A sounds quiet by comparison.
1. Why does normalizing the whole file to -16 LUFS make Host B's noise more audible, and what should the editor do instead?
Read the worked answer
Normalizing the whole file raises or lowers everything by one gain, so it cannot fix a level imbalance between the two hosts. If Host B is quieter on average, normalizing to hit -16 LUFS on the combined file pushes Host B's track up, and the -38 dBFS noise floor rises with it until it is audible between phrases. The editor should process each host separately: apply gain or compression to bring both voices to a similar average level, then treat Host B's noise with a noise reduction or expander stage before combining. After the voices are matched, the editor should measure the combined mix and adjust overall gain to reach -16 LUFS integrated, then check true peak against -1 dBTP. This keeps the loudness target from dictating how much noise is amplified.
2. What processing order should the editor use on Host B's track, and what risk does each step carry?
Read the worked answer
A sensible order is high-pass filtering to remove low rumble, then noise reduction or a downward expander to reduce the fan and room noise, then compression and level matching, then a de-esser if needed. Each step carries risk. Aggressive noise reduction can create artifacts and hollow the voice, so it should be applied modestly and checked on headphones. A downward expander can clip word tails and create pumping if the threshold is too high. Compression raises the noise floor along with the voice, so noise reduction should precede heavy compression. The editor should compare processed and unprocessed versions and prefer a lighter touch over an artifact-free but noisy track or a quiet but damaged one.
3. How should the editor verify the final deliverable meets the loudness and true peak targets?
Read the worked answer
The editor should measure the finished stereo mix with a loudness meter set to integrated LUFS and true peak, not rely on peak normalization alone. The integrated measurement should cover the full 48 minutes, because a short sample can miss loud or quiet sections. If the mix is below -16 LUFS, the editor raises overall gain; if above, lowers it, then rechecks true peak. If true peak exceeds -1 dBTP after reaching the loudness target, the editor applies a limiter or reduces gain slightly and remeasures, since limiting changes loudness. The editor should also spot-check the noisiest passages for audible artifacts and confirm the file's channel layout and sample rate match the platform spec before delivery.