Structure for the Ear

Structure for the ear is how a story is built when the listener has only sound. It starts from one fact about recording. A character is present for the listener while that character is speaking, or while someone else says the character’s name. At every other moment the character is absent.

On a stage or on screen, a silent body still takes up space. A person standing in the corner of a kitchen, saying nothing, watching, is a pressure on camera. On a recording that person is not in the kitchen and is not anywhere. The next time they speak, they enter the scene again, as though the room had never held them.

Three things follow from that. Characters have to be put back into the listener’s picture line by line. Entrances and exits have to be audible. Anything the eye would have carried has to be carried by a sound. The check that catches all of it is listening to the recording rather than reading the script, because the script keeps showing the writer people the listener cannot hear.

Core takeaways

• A character exists for a listener only while speaking or while being named. A silent presence does not register at all.
• Six speaking parts is the ceiling for a half-hour play written for strangers. Past six, the listener risks confusing them.
• The unit of writing is the sequence. It can be one line or one effect, and it lasts as long as its informative sounds last.
• Variety runs on four dials: sequence length, speaker count, dialogue pace, and location. They work as a revision check before recording and prove nothing when applied to a finished episode.
• Runtime is measured by reading the script aloud against a clock, with room left for effects and music. Page count and silent reading both fail.
• The sources behind these rules are graded differently. Some come from a broadcaster’s radio-play guide, some from practitioner sites and one studio blog, and some published percentages have no method attached and are not used as targets.

How many voices a listener can hold

The speaker cap for a half-hour play written for strangers is six. Past six, the listener risks confusing them. Six comes from a national broadcaster’s radio-play guide, and it is the only speaker number here that comes from the broadcaster. The guide was written for people who have never heard these voices before and will not hear them next week.

Per scene the working number is three or four. When a group cannot be avoided, one voice leads and names get said often. Three-to-four is practice, and it does not come from the same guide. Two independent practitioner clusters treat three-to-four as the field norm, two-handers as the workhorse, and five-plus as a scene that needs strong vocal contrast. That is still practice. It is not a measured finding, and it is not a citation to attach to the broadcaster.

The cap costs crowd texture. The full-room scene is the thing the number spends. A larger cast is staged across scenes instead of inside one. A wedding, a market, a staff meeting: those rooms exist as a sequence of smaller rooms, or as one room with one person talking and the rest named.

Character Voice is the other half of that ceiling: the dials that keep six voices distinguishable from each other. In audio there is no speech tag. A line that cannot be assigned to a speaker without a tag is a defect, not a stylistic choice.

Sequences and sound effects

The unit is the sequence. It can be one line. It can be a single effect. It lasts as long as its informative sounds last. The cut goes straight to the sound that carries information.

weak
FX: car draws up. Engine off. Door. Feet on gravel. Key in lock.

strong
FX: knocking at the door.
   Then the line that answers it.

The knock replaces the five-sound arrival. That is the effects budget applied. A listener who has already understood “someone is at the door” does not need the car, the gravel, and the key. Those five sounds are a film-school arrival played for an ear that cannot see the drive.

Cutting them costs the sounds that used to establish place. A hard-trimmed script leans instead on an ambient bed, the background wash that tells the listener where we are, because the footsteps and the key no longer will. Setting as Character is how that wash becomes pressure: place exists as a bed with acoustics that match the actors, and layering is how the pressure is established in sound.

Effects are used sparingly, for function or for mood. Excess is tedious. A door that slams every time someone is angry stops meaning anger.

Silence is a writing tool. Pauses let a listener absorb what just happened and prepare for what comes next. In a medium made only of sound, the absence of sound is an event the listener notices. A held breath after a name, or a room that goes still before a decision, is written on purpose and is not empty tape.

The four dials

Variety is what holds attention. It runs on four dials: sequence length, speaker count, dialogue pace, and location. The contrasts worth the most are a noisy many-voice stretch against a quiet monologue, and indoor against outdoor. A two-minute kitchen argument next to a twenty-second street crossing next to a one-voice room is the mechanism working. A script whose sequences run the same length at the same pace with three voices in the same room has all four dials at rest. No single sequence has to be badly written for the whole episode to go slack.

The dials cost a scheduling pass. Each sequence is read against the sequence beside it, the way a film cut is judged against the cut before it.

The dials cannot be falsified after the fact. A slack episode can always be described as a dial at rest. They earn their keep only as a revision instrument, used sequence against neighbour before the record button. Used that way they are a checklist. Used as a diagnosis of a finished episode they will confirm whatever the writer already believes.

Moving between scenes

Two transition rules do most of the work, and both come from a single self-published studio blog rather than from the broadcaster.

The first rule is that non-naturalistic conventions are taught early: a harp, a sting, a fade that does not occur in rooms. Teaching them spends early runtime.

The second rule is that the last line before a cut foreshadows the new location. The cut is seeded one line before it happens. “I’ll see you at the harbour” puts the harbour in the listener’s head a sentence early, so the new wash has somewhere to land.

A scene change arrives as a change in the wash under the voices. Place is re-established by layering: the wash first, a foreground detail second. A café is the room-tone of cups and talk, then one cup set down.

Acoustics have to match the actors. Interiors are reverberant and wet. Exteriors are flat and dry. A mismatched effect destroys the image the voices were building: a tiled kitchen with a dry outdoor footstep, a street with a bathroom slap. Perspective can flip by promoting a background sound to the foreground, so that rain heard as weather becomes rain someone is standing in.

Scene-change signalling is the complaint productions hear most often. Some productions answer it by giving each location a short sound-mark. That is practice as well.

Comedy

Audio comedy establishes a premise, escalates it roughly three times, and exits on the strongest line. The strongest line is spent leaving. The scene does not stay to enjoy itself.

A pause can be the punchline once the ensemble is established enough for the listener to fill the silence. If the listener cannot yet predict the character, the same pause is dead air. How much establishment is enough has not been settled.

Live-audience shows have a further mechanical demand. The script has to leave air for the laugh, or the next line is walked on.

Tonal Modulation holds the register-scale version of the same variety law: a beat lands in proportion to the contrast around it. The four dials are that law’s instruments on the audio surface. The exit-on-the-strongest-line shape sits beside that page’s timing rule for comic relief, which places the relief after the crest and never on the beat still working.

Timing the episode

The only reliable runtime is a spoken read against a clock, leaving room for effects and music. Page count will not produce it. A silent read will not produce it. The slot does not move. The pages do. One such slot is exactly 28 minutes excluding intro and credits, and it is where this discipline came from rather than a law for podcasts.

Each structural pass is another full spoken performance. That is what it costs to know whether the cut is shorter and whether the story is still there.

A podcast has no slot. The read-aloud still works without one. Which cutting disciplines survive when no external force sets the length has not been tested.

The industry default for a narrative podcast is three acts and a cold open. One major studio runs five acts and a cold open. Neither is a law. They are shapes listeners already know how to follow.

Openings and endings

A cold open primes mood, scene, stakes, and a question. Practitioners agree, without measurement, that most shows lose listeners in the first few minutes, and that starting slowly is the most reliable way to lose them.

The cold open costs an arresting beat. The beat that would have paid off later is spent early so the listener knows where they are and why they should stay.

The close mirrors the open. It ends on a question, a revelation, or a cliffhanger, so the listener starts the next episode owed an answer.

Where the rules stop holding

Six voices as a ceiling was a rule for a single half-hour play heard by people who had never met the cast. A serialized podcast with a returning cast and distinct actors breaks both conditions. The listeners already know the voices. The actors already sound unlike each other. Whether six still binds in that case has not been tested. Voice differentiation is the real limit, not a headcount. A serial that has already taught its listeners who is who can retire the cap and check the result by ear.

Complete fiction often runs about ten to twenty-five minutes. One daily serial has run about thirteen minutes since the early 1950s. Speech rates differ by language. Some shows combine a case-of-the-week with a serial frame. Those lengths are the clock a podcast writer actually cuts to. They are not the 28-minute slot, and they are not a measured finding. They are the shapes the slot does not supply: an episode that can be finished on a walk, or one that returns tomorrow at the same length.

The completion curve cannot show why listeners left. A dip may be caused by the topic rather than the shape. A stretch people leave because the subject went dead looks, on a chart, exactly like a stretch people leave because the sequences ran the same length.

The measuring instrument is the steepest cost of the method. A writer who cannot spend performance time cannot run the method as stated, and there is no cheaper substitute that still hears the room.

The cold listen and the drop-off timestamp

A cold listen is the episode heard without the script in hand. Every line has to be assignable to a speaker. Every character has to be listable. A forgotten presence is the failure. The test is not whether the writer remembers who was in the room. The test is whether a stranger, hearing it once, can name them. That is the same assignable-before-tag requirement Character Voice states as craft, made compulsory by a medium with no tags.

A sequence-cutting pass shows on the clock: a shorter read, and no story beat lost. If the read is shorter and a want, a reversal, or a name has gone missing, the cut was the wrong cut. The first production teaches the correction. Speak the pages, make the episode, measure the difference. The gap between the spoken estimate and the finished length is the allowance later drafts include from the start.

A drop-off timestamp is the minute listeners leave, consistently across episodes. A repair at the dip should move the dip or dissolve it. If the same minute keeps emptying after the structure around it has been rebuilt twice, the structure was not the fault. Industry benchmarks play no part in the check.

Chapter markers exist, and one large platform will auto-generate them from English transcripts when none are supplied. The lift figure attached to those markers has no stated methodology and is not used as a target.

What to change when a rule starts costing too much

When the read-aloud is eating the schedule, the next move is one full performance per draft rather than a pass on every line.

When the cast cap is starving the story, a serial retires the cap and verifies the result by cold listen.

When variety is being added that the scenes did not need, the draft is written without the dials, and the dials come back as a revision check.

When a drop-off has survived two structural repairs, restructuring stops. The content is what is left to re-examine.

How well each rule is sourced

Established, for the brief they were written for: a silent presence does not exist; six voices is the ceiling on a stranger-heard half-hour; a sequence is an informative sound and may be one line or one effect; variety runs on four dials; silence is a tool and effects are spare; runtime is a read-aloud against the clock; one broadcast slot is 28 minutes excluding intro and credits. Those are the load-bearing items from the radio-play guide. They were written for a competition slot and a listener who had never met the cast, and they are the best-sourced half of this material.

Practice, or a single blog, and graded that way: three to four speakers per scene; ambient layering and matching acoustics; comedy that escalates and exits; the two transition rules, which are to teach the convention early and seed the cut one line before; and an eight-item menu of fades, narration bridges, stings, conventional symbols, moving dialogue, and a moving microphone, with timings such as four-to-fifteen-second fades and three-to-five-word bridges. The menu and its numbers inherit the grade of one studio blog. They are not a law and they are not the broadcaster. A writer who treats the eight-item list as a required kit will spend early runtime teaching conventions the episode may not use.

Unmethodologized, and refused as targets: an 8% lift from chapter markers, a 60% floor for a compelling episode, a 70–80% band for an average. One platform documents that chapters exist. It does not publish those lifts. Rewriting the claim does not add data the sources do not have. A timestamp that moves after a repair is a check. A percentage copied from a trade post is not.

Three bounds stand against the method. The six-cap was written for strangers in a single half-hour. A writer without performance time cannot run the clock as stated. The completion curve cannot tell a structural fault from a subject people leave. None of the three is a reason to skip the cold listen.

The cap, the sequence, the cut, and the clock are all checked the same way. Listen to the recording, and name who is still in the room.

How to practice this

1. Read the script aloud against a clock, leaving room for effects and music. Notice whether the spoken length fits the slot. Do not use page count or a silent read.
2. Count the speaking parts in a half-hour script written for strangers. Notice whether the count is over six, and whether you can tell the voices apart.
3. Listen to the episode with no script in hand. Notice whether you can assign every line to a speaker and list every character.
4. Read each sequence against the one beside it before you record. Notice whether length, speaker count, pace, and location change between them.
5. Cut a sound sequence down to the one sound that carries the information. Then check the read is shorter and no want, reversal, or name went missing.
6. Find the minute listeners leave across episodes and rebuild the structure there. Notice whether the dip moves. After two repairs that do not move it, examine the content instead.

Related pages

• Character Voice: the cast ceiling and that page’s dials are two halves of one limit. In audio there is no speech tag, so assignable-before-tag becomes a requirement.
• Tonal Modulation: register-scale variety. The four dials are that law’s audio instruments, and the comedy exit sits beside comic-relief timing.
• Setting as Character: place exists as an ambient bed with matching acoustics. Layering is how that page’s pressure is established in sound.

Open questions

Whether the per-scene three-to-four cap, which survives only on secondary craft sites, should bind a serial the way the six-voice half-hour cap binds a one-off.

Whether the eight-technique menu and its timings, which come from one studio blog, earn a place as more than a checklist.

Whether six scales with runtime, so that a ten-minute episode and a fifty-minute episode ask for the same number of voices.

Which cutting disciplines survive when there is no 28-minute slot to cut to.

Whether any of the published completion benchmarks can be given a methodology.

How much establishment a pause needs before it is a punchline rather than dead air.

Sources

• Fiona Ledger, BBC World Service. How to Write a Radio Play (African Performance guide). Silent-presence law; six-character ceiling for a half-hour play; sequence as one line or one effect; four dials of variety; silence and spare effects; read the script aloud against the clock; 28 minutes excluding intro and credits. Mirrored at bbc.co.uk/worldservice and Lancaster teaching copies.
• “9 Great Techniques for Transitioning Between Scenes in Audio Drama,” weirdworldstudios.com. The two transition rules and the eight-item menu with 4–15s / 3–5 word figures. Single self-published studio blog.
• Rob Stearns and D. Price, greatnorthernaudio.com. Ambient beds, layering (wash then foreground), matching acoustics, wet interiors / dry exteriors, perspective by promoting background to foreground.
• thecomedycrowd.com, on audio-comedy shape (premise, escalate, exit). Jack Benny Program as the commonplace for pause-as-punchline; commonplacefacts.com is a weak cite for that tradition.
• lowerstreet.co and thepodcasthost.com. Cold open as mood / scene / stakes / question; three-act-plus-cold-open as the narrative-podcast default; one major studio (Airship / related) running five acts plus cold open. Practitioner consensus that the first minutes lose listeners.
• Apple creator documentation on podcast chapters and transcripts. Auto-generation of chapters from English transcripts when none are supplied is reported by industry press in 2025 (ireplay.tv and others). Apple documents that chapters exist. It does not publish an 8% completion lift.
• epicscribe.io and stage32.com as secondary restatements of a 3–4 per-scene practice. Not BBC-primary.
• podgagement.com / related completion-benchmark commentary. No methodology attached to the 60% / 70–80% bands used in trade writing.