From the booth. Side B, Issue 12.

The Room Does the Work.

— A note on room tone, and why we walk into hotel rooms with a microphone

The first thing we learned, after maybe three years of trying to make immersive audio, is that the room makes the recording. Not the microphone. Not the preamp. Not the post-production. The room. The walls, the carpet, the angle of the ceiling, the size of the air in the space. A great voice in a bad room is a bad recording. A modest voice in a great room is a recording the listener will play three times.

We have walked into roughly two hundred rooms with a microphone and a portable interface, looking for the rooms that have the tone. The best ones we have found, over those years: an old library in upstate New York with maple shelves and a slightly low ceiling; a hotel hallway in Lisbon at 4 a.m.; a basement apartment in Brooklyn that the owner had insulated with horsehair plaster. Each room sounds like itself before any human enters it. That is the sound we are after.

Most studio booths are designed to remove room. The reasoning is rational — predictable, controllable, identical from session to session. The result is what we call booth voice: clean, present, lifeless. A booth recording has no information about where the voice is. The listener hears a person speaking into nothing. The mental theatre is impossible to build because there is no room to build it in.

A great voice in a bad room is a bad recording. A modest voice in a great room is a recording the listener will play three times.

So we have built our practice around finding the room first, then bringing the voice to it. This is more expensive and more logistically painful than recording in a booth. It is the difference between the C3POE catalog and most published audiobook production.

A Eulogy for the Take We Didn't Use.

— On the second-best take and why it stays with you

Every session ends with one take we used and several we didn't. The takes we didn't use don't disappear. They live in a folder marked alternates, and a few of them haunt us for years. There is a Mary Oliver poem we adapted in 2023 where the narrator did the second take in a register half an octave below her usual range — quieter, more interior, almost a whisper. We didn't use it. We went with the take that had more clarity. The alternate take has stayed with me for three years. Some nights I think we made the wrong call.

Producers are taught to be decisive. Pick the take, move on, mix the record. But part of being a producer is being honest with yourself about which decisions you got wrong. The Mary Oliver session is one of mine. The published version is good. The alternate take is great. The difference is roughly the difference between Working tier and Reference tier in our own framework.

What I am learning, slowly, is to take the riskier take more often. The risk of the unusual register, the unconventional pace, the choice that makes the engineer uncomfortable. The takes we didn't use that haunt me are the ones that did something the published version was too cautious to do.

On the Half-Beat Hold.

— Why slowing the second-to-last sentence is the highest-leverage edit

If I had to teach a junior producer one technique that would lift the quality of nearly any recording, it would be this: slow the second-to-last sentence of each chapter by 15%. Not the last sentence — the second-to-last. The penultimate. The one that does the actual landing.

The reason: closing lines are easy to do badly because they call attention to themselves. Producers know they are important, narrators know they are important, the writing knows it is important. The result is that closing lines are often over-performed — too slow, too emphatic, too laden with the weight of the meaning. The penultimate sentence is the line that does the structural work — it sets up the close, it shapes the listener's expectation, it determines how the close will land.

Slow that line by 15%. Add an extra micro-pause before its final word. Let the listener arrive at the closing line already braced for it. The closing line does not need to do extra work; it just needs to land cleanly. Most recordings we have rescued in post had this single intervention — a half-beat hold on the second-to-last sentence — applied to every chapter close.

What AI Narrators Will Never Have.

— A producer's note on the technology that scared us in 2023 and stopped scaring us by 2025

We were scared of AI narration in 2023. By 2025 we had stopped being scared. The reason is specific: AI narration produces accurate readings of text, and accurate readings of text is not what we make. We make the texture of breath. We make the half-beat hold. We make the take that scared the producer.

The market for AI narration is large and growing — Paper 03 in this Volume has the numbers. AI is taking the high-volume, low-margin segments of the audiobook industry. We have not lost a single contract to an AI narrator. The contracts we win — premium short-form, literary, brand-funded immersive — are decided by qualities AI narration does not have: a real body breathing, a real choice made under pressure, a real human ambivalence that the listener registers without naming.

This is not snobbery. It is the simple mechanical fact that presence is a property of bodies. AI can simulate the surface of presence. It cannot have the thing that produces it.

Our advice to producers looking at this market: if you compete with AI on the accuracy axis, you will lose, because that competition is over. If you compete on the presence axis, you will not lose, because there is no competition on that axis. The trick is to stop trying to win on accuracy.

Presence is a property of bodies. AI can simulate the surface of presence. It cannot have the thing that produces it.

A Poem We Keep Returning To.

— On what makes a piece survive a tenth listen

There is a 90-second recording in our catalog from 2022 — a single poem read by a narrator we will not name here, in a hotel room in Edinburgh, with the heater clicking faintly under the recording. The poem is "The Layers" by Stanley Kunitz. The narrator read it nine times. We used the third take. The first two were too tight; the fourth and onward had begun to over-think the poem. The third take has something specific: a slight tremor on the line "how shall the heart be reconciled to its feast of losses?" that the narrator herself did not notice and cannot, when we ask her, identify or reproduce.

We have listened to that recording more than any other piece in our catalog. Producers, mixers, engineers, casting directors — we keep coming back to it. The recording is now four years old. It has not faded. If anything, the tremor has deepened.

This is the property we are trying to engineer: repeatability. Pieces that survive the tenth listen. Pieces that age into themselves. Most of what gets made in audio entertainment is built for a single listen — narratively closed, hooks-first, designed to deliver and move on. We are building, deliberately, the work that asks you to come back. The work that has more in it than you took on the first pass.

The Kunitz recording is the standard we use internally. When a piece is finished and we are unsure whether to release, we ask: will we be listening to this in five years? If the answer is yes, we release. If the answer is no, we go back into the studio.