Audio processing and loudness

0:00 Processing and loudness

Have you ever listened to a podcast in the car, then switched to another show and had to grab the volume knob because it was suddenly twice as loud? Or struggled to hear a quiet guest on a train, then been blasted when the host laughed? Those are processing and loudness problems. And they're fixable with a simple, consistent chain. In this lesson, you'll learn a six-step processing chain, what LUFS and true peak actually mean, Apple's published loudness recommendation, how to handle music and licensing, and a starting preset you can adjust by ear.

0:41 Why process?

Why does processing matter? Because people listen on earbuds, car speakers, laptop speakers and phones, often in noisy places. Processing makes voices clearer, more consistent and more comfortable across all of them. You don't need to be an audio engineer. A simple chain, applied gently and consistently, covers most shows. And consistency matters as much as any exact number. Listeners notice when one episode is much louder than the last, or when the guest is much quieter than the host.

1:16 Processing is photo prep

Here's an analogy. Think of processing like preparing a photo for print. First you remove dust spots. That's noise reduction. Then you adjust the color balance. That's EQ. You soften any harsh highlights. That's de-essing. You bring the shadows and highlights closer together so details show. That's compression. You balance two photos on the same page. That's leveling. And finally you set the overall brightness to the printer's standard. That's loudness normalization. Do each step lightly, and the photo looks natural. Overdo any of them, and it looks fake.

1:55 The six-step chain

Next, the chain, per voice track, in order. One, cleanup: light noise reduction, de-click and hum removal if needed. Two, EQ: a high-pass filter to remove low rumble, and gentle tone shaping. Three, de-essing, to tame harsh s and sh sounds. Four, compression, to even out loud and quiet phrases. Five, leveling, to balance the voices against each other. And six, loudness: normalize the final mix to a target, with a true-peak limit. Many editors offer one-click voice presets that combine these steps. Use them as a starting point, and adjust by ear.

2:35 Loudness targets

Now loudness. LUFS, loudness units relative to full scale, measures perceived loudness better than peak levels. Platforms normalize playback, so consistent loudness matters. A widely used podcast target is around minus sixteen LUFS integrated for stereo, or about minus nineteen for mono, with true peak no higher than about minus one. Apple's published recommendation is overall loudness around minus sixteen LKFS, with a tolerance of plus or minus one, and true peak not above minus one dBFS, measured using an international standard called ITU-R BS seventeen seventy. LUFS and LKFS describe the same measurement. Most importantly, measure your exported file, not just the project.

3:20 Example 1: loud host, quiet guest

A simple example. A host sits close to a dynamic mic. The guest recorded on a laptop mic, further away. The fix: light noise reduction on the guest track. A high-pass filter on both. A gentle EQ cut in the guest's boxy frequencies. Compression on both. Leveling so both voices sit at similar perceived loudness. Then final normalization to minus sixteen LUFS with a minus one limiter. And the most important fix of all: next time, the host sends the guest a USB mic and the setup guide.

3:58 Example 2: three shows, one standard

Now a realistic scenario. A network of three shows in the UAE, run by one small team, gets listener complaints that episodes sound different every week. Each editor uses their own settings. So they standardize. One processing preset per show, saved as a template. One mastering target: minus sixteen LUFS integrated and minus one true peak, applied with Auphonic as the final step. And a rule: every export is measured, and the number goes into the publishing checklist. Within a month, complaints stop, and switching between their shows no longer means reaching for the volume.

4:39 Watch me: a starting preset

Watch me build the starting preset from your lesson. Per voice track: high-pass filter at seventy to ninety hertz, lower for deep voices. Light noise reduction, stopping before voices sound watery. A small EQ cut in the boxy two hundred to five hundred hertz range, only if needed. A de-esser on the harsh five to eight kilohertz range, only when s sounds sting. A compressor at around two to one or three to one, with three to six decibels of gain reduction on louder phrases. Then on the master: a loudness target of minus sixteen LUFS integrated, and a true-peak limiter with a ceiling of minus one. I export, measure the exported file, and it reads minus sixteen point two. Within tolerance. Done.

5:32 Music and licensing

Now music and sound design. Short, on-brand intro and outro music, mixed below voices when they overlap, which is called ducking. Brief stingers to mark segments. And licensing is essential. You need rights to any music you use. "I only used ten seconds" is not a license. Options include royalty-free libraries with podcast and commercial licenses, commissioned music, or AI music tools whose terms permit commercial use, checked carefully. Keep license records. Avoid music under speech for long stretches, because it hurts clarity. And check that your music is licensed for video platforms too, because content-ID systems can flag unlicensed tracks.

6:16 Exports and video

Let's cover exports and video. Master archive: WAV, twenty-four bit, forty-four point one or forty-eight kilohertz. Podcast distribution: MP3, for example one hundred and twenty-eight kilobits per second mono for voice, or higher for stereo and music-heavy shows, or whatever your host recommends. Video: audio embedded per the platform's recommendations. Add metadata if your host doesn't. And use the same mastered audio for your audio and video versions, so listeners get the same experience everywhere. YouTube and Spotify normalize playback, so a clean, consistent master matters more than being loud.

6:55 Mistakes + recap

Here are the common mistakes. Heavy noise reduction that creates watery artifacts. Extreme EQ boosts. Over-compression that sounds squashed and tiring. Inconsistent loudness between episodes. Measuring the project instead of the export. And unlicensed music. Let's recap. Use the six-step chain, applied gently. Fix noise at the source when possible. Target around minus sixteen LUFS for stereo, with true peak at or below minus one. License all music and keep records. Export masters and distribution files properly. And check on headphones, laptop speakers and a phone.

7:32 Listening checks

One more practical habit: listening checks. Always listen on at least three devices: headphones, laptop speakers and a phone speaker. Check the loudest moment, usually laughter, and the quietest, usually a soft-spoken guest. Then compare the new episode's loudness and tone with your previous episode. If they sound like two different shows, something in the chain changed. These checks take ten minutes and catch most problems before listeners do.

8:02 Try this now

Here's your try this now. Build a processing preset in your editor using the six-step chain and the starting settings in the lesson. Process one episode. Measure the integrated loudness and true peak of the exported file. Then listen on three devices. Save the preset as a template for every future episode. In the next lesson, we'll package your episode with titles, show notes, chapters and transcripts.