Module-4.3-editing

Session 3

Why Some Podcasts Are a Pleasure to Listen To — Sound Problems, Real Listeners & the Filter Chain

Session 3 · Why Some Podcasts Are a Pleasure to Listen To — Ears First, Tools Second

What You Will Learn

  • How to hear and name the four most common sound problems: uneven levels, background noise, room echo and clipping
  • In which order to repair them — and why levels always come first
  • Where and how people really listen to podcasts, and what that means for the mix
  • The minimal 4-step filter chain for a spoken voice: high-pass, noise reduction, compression, EQ
  • How to judge every processing step with A/B listening — and recognise over-processing
  • The difference between peak level and LUFS, the loudness number the platforms care about

📋 Session info

⏱ Duration: about 3.5 hours, including a 75-minute chain-building exercise

👥 Group: 10 participants

🌐 Language: English

🔗 Builds on: Sessions 1–2 — the editing session and the cleaned interviews

🔧 Hands-on: everyone builds the filter chain on their own recording and levels the two speakers

Train the ear before you train the hand.

Before touching a single filter, learn to hear what is wrong and to name it. Then meet the real listener — on a bus, in a kitchen, in a car — and build a short chain of four processing steps that makes a voice clear and steady for exactly that listener.

Name What You Hear

The session opens with ears, not tools. Play five pairs of short samples on a room speaker — each time the problem version first, then the fixed one — and ask the group to name what they heard, first in their own words, then with the proper term.

⚖️ 1 · Uneven levels

One voice loud, the other quiet. The listener keeps reaching for the volume.

💨 2 · Background noise

Hiss, hum, traffic or wind underneath the voice.

🏛️ 3 · Room echo

Reverb from hard walls: the voice sounds distant, as if in a bathroom or a hall.

💥 4 · Clipping

Harsh distortion from an input level set too high.

The fifth pair is a clean reference, so that the group hears what "good" sounds like, not only what "bad" sounds like. Then ask one question: which of these would make you switch off a podcast fastest?

The number one complaint

Inconsistent volume between speakers. A practical rule of thumb: keep the loudest and the quietest voice within about 2–3 dB of each other.

The 2–3 dB guideline is a working rule used by audio practitioners, not an official standard — but it is a useful target. It also explains the order in which problems should be repaired:

First

Levels

Balance the speakers.

Then

Noise and echo

Clean the signal.

Last

Cosmetics

Polish the tone.

Polishing an unbalanced recording is polishing the wrong thing. If you collected the group’s answers about their favourite podcasts in Session 1, bring that sheet back now: "the hosts are equally loud", "no background hiss". As listeners, they already knew everything this session teaches them as producers.

🎓 Trainer Tip: Keep every sample under 20 seconds — the exercise lives on quick contrasts. You can build the pairs from your own practice recordings: a raw outdoor clip and a cleaned version for noise, a clip from an echoey room for reverb, and two clips artificially unbalanced or overdriven for levels and clipping. Test the room speaker at conversational volume the evening before; this session depends on whole-group listening.

🎙️ Audio note — to be recorded and embedded

Topic: Five sound problems — a listening test

Suggested script outline:

  • Uneven levels: two speakers, one much quieter — then balanced
  • Background noise and room echo: each as recorded, then cleaned
  • Clipping: a distorted sentence, and why it cannot be repaired afterwards
  • A clean reference fragment, and a short spoken reminder of the repair order: levels, then noise and echo, then cosmetics

[Replace this block with a WordPress audio block after recording]

✏️ Exercise — click or tap to expand ▼

▶ ✏️ Try This With Your Group

Name What You Hear

Training the ear to recognise and name sound problems. Five prepared pairs of short samples (problem, then fixed), a room speaker and a flipchart.

Step 1 — listen (10 minutes): play each pair. After each one, the group describes the problem in their own words; write the words down.

Step 2 — name it (5 minutes): next to each description, add the proper term: uneven levels, background noise, room echo, clipping, clean reference.

Step 3 — rank (5 minutes): a show of hands — which problem would make you switch off fastest? Then introduce the repair order.

⏱ Time: about 20–30 minutes

📋 Alternative — Take-home task: Ask everyone to listen to three different podcasts and note any of the four problems they hear. Which one bothered them most, and in which programme?

Where and How People Actually Listen

It is easy to imagine the listener in a quiet room with good speakers. The real listener is somewhere else entirely — and that changes how a podcast should be mixed.

📱 The phone with headphones

The dominant way of listening: while commuting, cooking or walking the dog. The environment is noisy and attention is shared.

🚗 The car

A major listening room, with engine and road noise underneath everything.

🔊 Smart speakers

Often a single small mono speaker across the kitchen.

🎧 Wireless earbuds

Short sessions, frequent interruptions, and an instant volume comparison with whatever played before.

Podcasts are also increasingly watched, not only heard. According to Edison Research’s report The Podcast Consumer 2026 (US data, published June 2026), 82% of weekly podcast consumers in the United States watch video podcasts, while 78% listen to audio only — many do both. Video is beyond the scope of this module, but it shows that the format is evolving and the audience is broader than the classic "audio person".

What this means at the editing desk

📲 Check on cheap gear

Good monitors flatter; the phone tells the truth. A phone test before publication is not optional.

🔈 Think mono

Smart speakers and a single earbud play in mono. Spoken word rarely needs stereo tricks anyway.

📏 Stable loudness

A listener on a bus cannot ride the volume knob for you. Steady loudness beats everything.

For older listeners, clarity comes first

Clear speech and consistent loudness matter far more than any production flourish. Everything in this module serves exactly that.

🎓 Trainer Tip: If someone asks whether their organisation should make video podcasts, acknowledge the trend honestly — the numbers are real — and park the topic: video production is a separate subject. Clean audio transfers one to one to any future video format, so nothing learned today is wasted. Statistics like these change every year; check for the latest edition of the report before quoting it.

✏️ Exercise — click or tap to expand ▼

▶ ✏️ Try This With Your Group

Where Do Our Listeners Listen?

Replacing the imagined listener with the real one. Sticky notes, a wall or flipchart, and one short podcast fragment.

Step 1 — map (5 minutes): everyone writes on sticky notes where the older people they work with listen, and on what device. Stick them on the wall and group them.

Step 2 — test (5 minutes): play the same fragment on a good speaker, then on a phone’s own speaker from across the room. What got lost?

Step 3 — conclude (5 minutes): agree on three mixing rules for your own audience.

⏱ Time: about 15 minutes

📋 Alternative — Take-home task: Ask everyone to ask three older people they know where and on what they listen to radio or podcasts, and bring the answers to the next meeting.

The Filter Chain: The Theory Minimum

Four processing steps take a cleaned voice recording from "recorded" to "a pleasure to listen to". They are applied always in the same order, and always checked with your ears.

Step 1

High-pass

Removes rumble.

Step 2

Noise reduction

Removes hiss, gently.

Step 3

Compression

Steadies the volume.

Step 4

EQ + de-esser

Polishes the tone.

1 · High-pass filter

Cuts the deep rumble a voice does not need: air conditioning, desk thumps, traffic bass, leftover wind.

Sane range: a cutoff around 80–100 Hz for speech — an orientation, not a dogma; lower for deep voices, higher for high ones.

Too much sounds like: a thin, telephone-like voice.

2 · Noise reduction

Removes steady background hiss and hum. The key word is gently.

Sane range: the lowest setting that helps.

Too much sounds like: a metallic, watery, underwater voice.

3 · Compression

The automatic hand on the volume knob: it pulls loud moments down and lets quiet ones up, so the voice stays steady on a noisy bus.

Sane range: a ratio between 2:1 and 4:1 for speech.

Too much sounds like: pumping, breathless, tiring speech.

4 · EQ + de-esser

The finishing touch: reduce muddy low-mids if the voice sounds boxy, add a little presence if it sounds dull, and tame sharp "s" sounds. This is also the proper fix for popping "p" and "b" sounds.

Sane range: small moves, 2–3 dB at a time.

Too much sounds like: a harsh, artificial "radio DJ" tone.

A de-esser is a processor that turns down only the sharp "s" sounds. Not every editor has one built in; Hindenburg, for example, does not include one as standard. Where there is none, a gentle EQ cut in the range where the "s" sounds sit is a workable approximation, and a de-esser plugin can be added later.

Rules that outrank every setting

🔁 A/B after every step

Toggle the effect on and off. If you cannot hear what it did, it should do less, or nothing.

🤶 Less is more

Every processor is also a distortion. The chain serves the voice, not the other way round.

⚖️ Levels come first

Before any filter, bring the speakers within about 2–3 dB of each other.

Why the order matters. Noise reduction works better on a signal without rumble. Compression amplifies whatever noise is left, so you clean first and compress after. And EQ judges the voice you actually have after the previous steps — so it comes last.

💡 Tip: A little noise with a natural voice beats total silence with a robot.

🎓 Trainer Tip: Keep the theory short — one sentence per step — and demonstrate the noise-reduction overdose live: crank it far too high and let the group hear the metallic, underwater voice. Learners remember that sound far better than any number. Before the workshop, check where the effects live in your editor’s current version, so that you can name them precisely.

Build the Chain, Step by Step

The best way to learn the chain is in synchronised steps: the trainer performs each step on the projector, and everyone applies it to their own material — the interview cleaned in Session 2, or another raw recording. After each step, two minutes of A/B listening before anyone moves on.

1

High-pass on

A/B: what disappeared? Usually a rumble nobody had noticed.

2

Noise reduction

Gentle. Then overdose it for 30 seconds on purpose, and dial back.

3

Compressor

Ratio 2:1–4:1. A/B on a laugh or an energetic moment.

4

EQ (optional)

For those ahead. Others skip without guilt.

5

Check the levels

Interviewer and interviewee back to back: within about 2–3 dB?

The first three steps carry most of the value. The level check deserves a word of its own: Hindenburg automatically levels every clip when it is imported, using a real loudness analysis. So the task is not to level from scratch but to listen to the two speakers back to back, confirm by ear that they sit close together, and correct the clip volume only where your ear disagrees. Understanding what the tool did for you — and when to overrule it — is the actual skill.

Challenges for fast finishers

🏛️ The echo recording

What can EQ do against room echo, and what can it not? The honest answer: not much — echo is baked into the signal. AI tools, covered in Session 4, do better.

🌳 The outdoor recording

Find the exact point where noise reduction starts eating the voice.

Decide with your ears, not the numbers

Close your eyes, toggle the effect, listen to 20 seconds.

🎓 Trainer Tip: Participants tend to set everything by eye — by the numbers on the screen — instead of by ear. Keep repeating the same sentence while you circulate: "Close your eyes, toggle the effect, decide with your ears." The deliberate overdose in step 2 is worth the 30 seconds: hearing the failure once inoculates against it for good.

✏️ Exercise — click or tap to expand ▼

▶ ✏️ Try This With Your Group

Build the Chain

Applying the four steps to your own recording and hearing what each one does. Laptops with the editor, each participant’s cleaned interview, headphones and the Filter Chain card.

Step 1 — high-pass and noise reduction (25 minutes): follow the trainer, step by step. After each, A/B for two minutes. Overdose the noise reduction on purpose, listen, then dial back.

Step 2 — compression and optional EQ (25 minutes): add the compressor and A/B on an energetic moment. Those ahead try a small EQ move.

Step 3 — levels (15 minutes): listen to the two speakers back to back and adjust the clip volume until they sit within about 2–3 dB. Fast finishers take on the echo and outdoor challenges.

⏱ Time: about 75 minutes

📋 Alternative — Take-home task: Ask everyone to apply the first three steps of the chain to a phone recording at home, with A/B listening after each. Bring the before and after versions, and one sentence on which step made the biggest difference.

Before and After — and What Is LUFS?

By this point, the difference between a raw recording and the cleaned, processed version is usually dramatic. Let two or three volunteers play a fragment on the room speaker: first the raw version, then the current state. Hearing the progress on a shared speaker makes the whole morning tangible.

Then comes one concept the final session builds on. There are two different questions you can ask about loudness:

📈 Peak

How loud is the loudest split second? This tells you whether the file risks distortion.

👂 LUFS

How loud does it feel to a human ear over time? This is what listeners notice — and what the platforms measure.

A file can have safe peaks and still feel whisper-quiet, or feel crushingly loud at the same peak level. That is why platforms care about LUFS. A quick way to see it: export a short fragment and drop it into a free online meter such as the Youlean Loudness Meter, which shows the measured value. Hindenburg itself works with these numbers: it levels clips on import and can export at a loudness target. Which number to aim for is the subject of Session 4.

💡 Tip: Keep the LUFS introduction to concept only. One idea — "how loud it feels over time" — is enough for now.

🎓 Trainer Tip: Test the meter workflow end to end before the workshop — export from the editor, drag the file into the browser — so the demonstration takes one minute, not ten. End with a cliffhanger rather than a lecture: "In the next session you will learn which number the platforms actually want, and you will hit it before you go home."

✏️ Exercise — click or tap to expand ▼

▶ ✏️ Try This With Your Group

Raw versus Ready

Celebrating progress and meeting the loudness number. Volunteers’ recordings, a room speaker and a free online loudness meter.

Step 1 — play (10 minutes): two or three volunteers play the same fragment raw, then processed. The group says what improved.

Step 2 — measure (5 minutes): export both versions of one fragment and measure them in the online meter. Compare the two numbers.

Step 3 — ask (5 minutes): did the version that felt louder also measure louder? Introduce the difference between peak and LUFS.

⏱ Time: about 20 minutes

📋 Alternative — Take-home task: Ask everyone to measure the loudness of two episodes of different podcasts in a free online meter and bring both numbers to the next meeting.

Key Takeaways

  • Ears first. Learn to hear and name the four common problems: uneven levels, background noise, room echo and clipping.
  • Repair in order. Levels first, then noise and echo, cosmetics last.
  • Balance the speakers. A practical rule of thumb: the loudest and quietest voice within about 2–3 dB.
  • Mix for the real listener. Phones, cars, smart speakers and cheap earbuds — check on cheap gear, think mono, keep loudness stable.
  • Four steps, one order. High-pass, gentle noise reduction, compression at 2:1–4:1, and a small EQ touch.
  • A/B after every step, and less is more. If you cannot hear what a step did, it should do less, or nothing.
  • Peak is not loudness. LUFS measures how loud a file feels over time — the number the final session turns into an export target.

Resources & Downloads

A printable one-page card with the four steps, their sane ranges and the warning signs of over-processing. Keep it next to the laptop whenever you process a voice.

  • 📄 The Filter Chain — the 4 steps in order, what each does, the sane ranges, the A/B rule and what too much sounds like

Sources and tools mentioned on this page