Caption a walkie-talkie channel for about $24

The one thing on set nobody has solved

Earpieces assume you can hear. Speaker mics assume you can hear. On a film set, in a warehouse, at a race pit or on a construction site, the radio is how the day is coordinated — and there is no accessible version of it.

Relaying is not a solution. It makes another person into your infrastructure, it lags behind the conversation, and it puts you in the position of constantly asking for help to do a job you are perfectly capable of doing.

Playing the radio out loud is not a solution either. On a set it is disruptive during a take, and on most channels the traffic is not for the room to hear.

So: wire the radio directly into your iPhone. It is silent to everyone around you, the transcript stays on your device, and your wrist taps when your name goes out over the air.

What we tested, and how

We built four different ways of getting radio audio into an iPhone, then read the same paragraph over the air into each one and scored the captions word by word against the original.

The test paragraph was 44 words: "The 10 watt rating helps at the margins, but a 10 watt handheld doesn't go twice as far as a 5 watt one — maybe 20 to 40 percent farther in fringe conditions. The taller antennas in your kit matter more than the extra watts."

It is deliberately awkward: repeated technical terms, numbers, percentages, and a couple of words that sound like other words. Radio audio is narrow and compressed, so this is roughly the hardest kind of sentence to get through it intact.

One read-through per setup. That is enough to separate a good rig from a bad one, and not enough to call it a laboratory result — we would rather tell you that up front than dress up four samples as science.

What the captions actually looked like

The winning setup, at 91%, got the entire first sentence perfect including every number. It kept the word "twice", and it was the only setup that got "fringe conditions" right. Its errors were cosmetic: "farther" became "further", and the last sentence lost "taller antennas".

The $38 setup scored almost identically at 89% but made a far more expensive mistake. It dropped a single word — "doesn't go twice as far" became "doesn't go as far", the opposite claim, stated confidently.

That is worth understanding before you trust any caption system: a low error count can still carry a serious misunderstanding. Error count and error cost are different things.

The single-cable setup at 73% failed in the most revealing way. "Watt" became "block" all three times — the one word the paragraph was actually about. When a setup is starved for signal, the words it loses are the content words, not the filler.

Why the cheapest setup won

This surprised us, so it is worth explaining properly. Audio chains are subtractive. Every component between the radio and your phone can lose detail or add noise, and nothing further down the line can put back what an earlier stage dropped. The fewest, cleanest stages win — almost regardless of what each stage costs.

The single-cable setup looks like it should win on that logic, since it is one physical cable. But it hides three separate gain stages inside itself: the radio's volume knob, an inline volume control, and its own built-in converter chip. We had to turn all of them well up to get a usable level, and each one amplified hiss right along with the speech.

The RØDE AI-Micro is a genuinely good piece of equipment, built to be forgiving about unpredictable input. But the $9 Apple USB-C adapter contains the same class of converter, and Apple tuned it well. Feeding either one a steady, already-compressed line from a radio, the expensive one's extra capability simply goes unused. Fancy hardware helps when the source is difficult — and a radio's line output is not difficult. It just needs to be left alone.

The one clever part

There is a reason the winning setup needs only two parts, and it explains a mistake many people make first.

An iPhone captures audio from the microphone pin of a headphone connection. A radio puts its audio on the headphone pins. They do not line up. This is why the obvious attempt — a basic adapter, an aux cable, and an Apple adapter — produces silence. The audio arrives at the phone on pins the phone is not listening to.

Something in the chain has to move the signal across to the microphone pin. A push-to-talk adapter already does it internally, because it is built to accept a phone-style headset in the first place. It remaps the pins and gives you a microphone and a transmit button, all in one part costing about $7.

That is the whole trick. One part doing two jobs is why the recommended setup is both the cheapest and the most accurate.

What you need

Two parts, about $24 total, plus an iPhone with a USB-C port.

Check your radio's connector first — this is the easiest way to buy the wrong part. Baofeng and most consumer radios use a Kenwood-style two-pin plug, which is what the adapter above fits. Motorola production radios like the CP200d use a Motorola "M1" two-pin plug: it looks similar and does not fit. Newer digital Motorolas use a multi-pin connector instead. The Motorola section below has the details.

Using it with Hearing Buddy

Once the audio is in, the app works the way it always does — the radio is just another thing Buddy listens to. A few features earn their keep more than usual here.

Nothing is ever played out loud. The radio's speaker is disconnected the moment you plug in, and Buddy only listens to incoming audio — it never plays it back. Everything stays on your device. Channel traffic on a production or a work floor is often confidential, and it stays that way.

Five minutes of radio settings that buy real accuracy

Most of the remaining caption errors come from the radio, not the phone. These help:

• Use wide bandwidth (25 kHz) rather than narrow (12.5 kHz) if your radios support it. More high-frequency detail survives the trip, and high-frequency detail is exactly what tells "fringe" apart from "French".
• Set squelch so an idle channel is completely silent. Open-channel hiss is a constant stream of noise into the captions and can invent words out of nothing.
• Turn the roger beep off. That end-of-transmission tone is a loud non-speech burst landing exactly where a sentence ends.
• Turn VOX off. On any setup with a transmit path, voice-activated keying can put your room on the air without you touching anything.
• Stay on the same band as your crew. Most workplaces run UHF because it handles walls and interiors better.

You can talk back, too

The push-to-talk adapter carries a small microphone and a transmit button alongside the audio it feeds your phone. Press the button, speak, and your voice goes out over the channel exactly as it would from any standard earpiece kit.

Then something unexpected happens: your own transmissions show up in your transcript. The adapter's microphone feeds the phone as well as the radio, so what you said lands in the same thread as everyone's replies.

That closes a loop that has never really been closed before. If you voice on the channel but cannot hear the response, you have historically had no way to confirm what actually went out — you say something into a device and hope. Now you can see your own words and the reply to them, in order, on one screen.

If you do not voice on the channel, none of this gets in your way. The setup receives perfectly well on its own.

If your radio is a Motorola

The Motorola CP200d is the film and TV standard — it is what rental houses stock by the case. It uses a Motorola two-pin plug, known as M1, which is physically different from the Kenwood-style plug on consumer radios. The adapter listed above will not fit it. M1 also covers the CP100d, CP125, CP150, CP185 and DTR series.

Everything else in this tutorial is identical; only the first part changes. The one to start with is a Pdflie M1 push-to-talk cable with a 3.5 mm female jack, around $13 — the same shape of part as the Kenwood-style adapter that won our bench test: a two-pin plug for the radio, a transmit button, and a jack built to take a headset.

Buy one more thing alongside it, though, and here is the honest reason why. A headset jack normally does the opposite of what you need: it sends audio out on the headphone pins and listens on the microphone pin, because it expects earbuds with a mic attached. An iPhone records from that same microphone pin. Two devices both listening on one pin and nothing driving it is silence — which is exactly why the obvious build with a plain adapter and an aux cable does not work.

Our $24 Kenwood build works anyway, so that adapter is getting receive audio onto a pin the Apple adapter can read. We think headset-style push-to-talk adapters remap the pins internally to make that happen. That is a reasonable explanation drawn from one working sample, not something we have measured — and the M1 cable is a different manufacturer, a different connector and a different board inside.

So until someone runs it on a real CP200d, order a RØDE SC7 patch cable, about $20, at the same time. Its entire job is to put line-level audio onto the microphone pin of a 3.5 mm jack. Try the two parts on their own first. If the Input Level meter stays flat while someone is talking, put the SC7 between the M1 cable and the Apple adapter and it will work. Worst case the Motorola kit costs about $42 instead of about $22 — cheap insurance against a shopping list that fails on the most common radio on set.

If the production runs newer digital radios instead — XPR 3300e, XPR 3500e, XPR 6000 or 7000, or APX — those use a multi-pin side connector rather than M1, and need a multi-pin-to-two-pin adapter before anything else. That is an extra part and an extra thing to go wrong. Borrowing a CP200d is usually easier.

One tip worth more than any of the hardware: ask whoever runs comms for a spare radio rather than modifying your own. Every production has a bin of them. A dedicated radio wired to your phone means your working earpiece stays untouched, and if the caption setup ever misbehaves you unplug one thing instead of re-rigging in the middle of a shot.

To be completely clear about what is tested and what is not: the roughly $24 two-part build was bench-tested on a Kenwood-style consumer radio and scored 91%. The Motorola path has not been tested on a CP200d yet. The parts are right and the principle is the same, but the SC7 is there because we would rather you have a working setup than a confident recommendation.

When something is wrong

Almost every problem is one of five things, and four of them are the volume knob.