Getting Sound Into the Frame

Sound is the part we do not control

The frame turns sound into light. Everything on the light side of that sentence is ours — the panel, the plexi, the software, the way the circles move. Everything on the sound side arrives from outside, from equipment we did not design, in a room we have never seen.

That asymmetry is the whole problem. A customer might be playing music from their phone into wireless earbuds, or through an Apple TV into a television, or from a laptop, or through a Bluetooth speaker on a shelf, or at an opening with a proper sound system. Each of those is a different piece of engineering for us, and some of them are not engineering at all — they are a licence agreement with Apple, or a fee to a standards body, or a wall we cannot climb.

Here is the uncomfortable version. The most common way people listen to music in 2026 is the one we can least reach. A phone playing to wireless earbuds produces no sound in the room and no electrical signal anywhere we can touch. There is nothing to plug into and nothing to hear. No amount of cleverness fixes that, and we should say so out loud early rather than discover it in a customer's living room.

Everything else is more tractable than it first appears — often much more, and usually because somebody else has already paid for the hard part. This document lays out which situations we can serve, what each one costs, and the three decisions we need from you before the first piece is wired.

Two paths that look like one

There are two completely different jobs hiding inside the phrase "getting audio to the frame", and confusing them is how this kind of product goes wrong.

The first is the analysis path: sound coming in so the frame can make light. It needs a loudness value and a rough sense of which frequencies are busy — sixteen numbers, sixty times a second. Mono is fine. Background noise is fine. A cheap converter is fine. A fiftieth of a second of delay is invisible to the eye. It is an extraordinarily forgiving job, and almost anything satisfies it.

The second is the listening path: sound going out to the customer's ears. One click, one dropout, one second of silence at a dinner party, and the product is defective no matter how beautiful the light was. This is hi-fi engineering, with hi-fi expectations and hi-fi returns.

The diagram states the recommendation. The frame should sit on the analysis path only. It listens to the music; it does not carry it. If the frame crashes, the lights go dark and someone raises an eyebrow. If the frame were carrying the music and it crashed, the party stops and the box comes back to us. That is not an engineering difference — it is the difference between selling a light and selling a piece of hi-fi equipment, with all the expectations that come with the second.

The second framing point follows from the first. The instinct is to try to capture what a phone is playing. That is the hardest possible version of the question and on an iPhone it is simply not possible — Apple has never allowed one app to read another app's sound, and there is no workaround, no entitlement, no back door. Turn it around and it becomes easy: every phone already has a one-tap way to send its music somewhere else. If the frame appears in that list, the customer picks it and clean audio arrives with no trickery at all.

What the others built, and what it cost them

This category has been around long enough to have a graveyard, and the pattern in it is unusually clear. Every audio-reactive light product still on sale gets its sound one of two ways: a microphone, or a tap on a wire the customer already has. The products that tried to be clever about accepting any source are all discontinued.

1. Philips is the cautionary tale. Their Sync Box sits in the middle of the customer's television setup and reads the video going past. It works, it costs around €350, and its reviews are dominated by television problems rather than lighting ones — black screens, flicker, copy-protection errors, televisions switching themselves off. It took them five years of firmware updates to get input switching right. And it still cannot see the television's own Netflix app, which is why they eventually paid to put a €130 app inside Samsung and LG televisions instead. a company a thousand times our size, five years, and they replaced their own flagship with software
2. Nanoleaf ran the experiment for us. Their first sound module had both a microphone and a cable input. When they redesigned, they dropped the cable and kept the microphone. Their €99 camera-based television kit is judged by reviewers to react about as well as the €240 competitor that sits in the HDMI chain. the simplest input won on the second iteration, at the company that had tried both
3. Govee, the volume leader, ships a microphone and one button. No selection, no pairing, no app required. It reacts to whatever is in the room and it survives being switched off at the wall. the highest-shipping answer in the category is also the cheapest one

None of this means we should copy them. It means the expensive routes have been tried by people with more money than us, and the results are public. Where we can be better is not in accepting more sources than Philips — it is in the piece itself, and in being honest about which rooms it works in.

The build

At twenty hand-built pieces the calculation is completely different from a product with a factory behind it, and it is different in our favour. We do not buy licences. We buy finished boxes that already have them. Somebody else has paid Apple, Spotify, Google and the Bluetooth standards body, passed the certification, and sells the result for a hundred and fifty euros. Our job becomes wiring rather than negotiation.

One box removes most of the difficulty

A WiiM Pro, around €150, is a certified receiver for AirPlay 2, Google Cast, Spotify Connect, TIDAL Connect and several others, with ordinary audio outputs on the back. The customer opens Spotify, or the AirPlay menu on their iPhone, sees the frame by name, and taps it. No agreement with Apple, no non-disclosure agreement, no per-unit royalty, no software of ours to maintain.

That single part answers what would otherwise be three separate closed doors, each requiring a corporate relationship we are too small to obtain. It is the best value in this entire document.

Bluetooth without ever talking to the standards body

Buy a finished receiver — an Audioengine B1 at around €190 is a good one, and suits a piece that should be beautiful inside as well as out. Cheaper receivers do the same job for €60 where that does not matter. Because it stays Audioengine's product rather than becoming ours, the €12,000 qualification fee that would apply to a Bluetooth product of our own does not arise. The one rule to respect: we must not print the word "Bluetooth" on our own packaging, because that specifically requires membership and certification. "Works with your phone" says the same thing.

Television sound, solved by a switch on a thirty-euro box

A small HDMI audio extractor sits between the Apple TV and the television, passes the picture straight through, and gives us the sound. The ones with a 2CH switch do something quietly brilliant: they tell the Apple TV that only ordinary stereo is supported, so the Apple TV sends ordinary stereo. Two things follow. We never have to decode Dolby, which would cost tens of thousands a year in licensing. And the customer never has to find a settings menu, because the negotiation happens between the two boxes without them.

Everything in the middle column is a finished product somebody else certified, and everything on the right is ours. The prices are real and current. Nothing in this drawing requires a licence agreement, a legal opinion, or a conversation with a company that will not return our calls.

The most important thing about this arrangement is that it is modular per commission. Every frame gets the interface, the microphone and a line-in socket — around €160. The WiiM, the Bluetooth receiver and the HDMI extractor are options, fitted when the room needs them. At €150, €190 and €40 those are decisions we can make per piece rather than per product line, and that freedom is precisely what building twenty of something buys you.

Choosing what it listens to

With several sources connected, something has to decide which one the frame is listening to. The obvious answers are a selector switch or a knob. The better answer needs no switching hardware at all.

A four-input audio interface presents all four inputs to the computer at once, as one continuous stream. So the frame does not switch — it listens to everything, measures how much music is on each input, and follows whichever one is playing. A threshold, a few seconds of patience so it cannot flicker between two, and that is the entire mechanism.

Three things fall out of that, and they are all good.

• Nothing mechanical. No relay to click, no switch to wear out, no contact to corrode in ten years.
• The frame knows what it is hearing. Which means it can say so — on the phone, and in the light field itself. A brief unmistakable pattern when the source changes, then back to the music. We already own hundreds of addressable circles; putting a status light on a light sculpture would be absurd.
• It can cross-fade instead of cutting. Because the frame is comparing rather than switching, a change of source can be a dissolve rather than a jump. On a piece like this, that is the more beautiful behaviour, and it costs nothing extra.

The interface's own gain knobs — one per input — get set once at build time to match each source, then never touched again. That is exactly the kind of setting the software already treats as physical: adjusted with the piece in your hands during installation, and deliberately out of reach of anyone holding a phone.

If a physical control is wanted for the object's own sake, one rotary knob works well as an override rather than as the primary mechanism — turn it to pin a source, push it to release. But it should earn its hole in the plexi on design grounds, not because the electronics need it. They do not.

The things we cannot do

Four honest limits. Three are walls; the fourth is a choice we should make deliberately.

1. We cannot reach someone listening on wireless earbuds. A phone sends its audio to exactly one destination at a time, and the room stays silent. There is no signal to tap and no sound to hear. The only thing available to us is the information that music is playing — the track, the artist, whether it is paused — which Apple does make available over a short-range wireless link without our touching the audio. That would let the frame breathe with the piece rather than react to it, and let a knob control the phone's volume. It is worth prototyping, but it is a different kind of feature and we should not pretend otherwise. a wall, with a small consolation prize behind it
2. We cannot read what an iPhone is playing. No interface exists, none has ever existed, and no app of ours could be approved to do it. Android technically allows it, but Spotify switches it off and the capture stops the moment the phone screen locks. This should be considered closed rather than difficult. a wall, with nothing behind it
3. We cannot take audio out of an Apple TV without a small extra box in the chain. The picture is encrypted, and the licence to decrypt it costs tens of thousands a year and obliges us to keep secret keys safe inside a computer that also runs a browser — which we could not honestly claim to do. The €30 extractor is not a workaround, it is the correct answer: someone else holds the licence and the customer can replace it without touching our piece. a wall, with a thirty-euro door in it
4. We should choose not to be the customer's hi-fi. We could put the music through the frame and out again, and gain volume control and the ability to mix. We would also gain a boot delay before the music starts, and a failure mode where our software crashing means their music stops. That trade is bad for us. It is worth saying plainly that this is a decision rather than a limitation — if we ever want to sell the frame as a speaker, everything changes, including the price. a choice, and the most consequential one in this document

What it costs and what it buys

The parts, per piece, at the numbers we would actually pay:

Set against what the same capabilities would cost if we built them ourselves rather than buying them in: around $13,000 in fees before the first Bluetooth unit ships, a further $15,000 for the audio format Apple phones prefer, $20,000 to $35,000 a year for the HDMI and copy-protection licences, and an agreement with Apple whose terms we would only learn after signing and waiting months. None of that is money we would spend at twenty units; all of it is capability we get for the price of a few hundred euros of hardware.

"The cheapest engineering decision available to us is to let other companies keep paying for the difficult part."

What we need from you is three decisions, and only the first is urgent.

One: is the frame ever in the listening chain? Everything in this document assumes it is not — that it listens, and the customer's music goes wherever it was already going. That keeps us out of an entire category of returns and keeps the object a light rather than a stereo component. If the answer changes, the price, the parts and the risk all change with it.

Two: do the bought-in boxes go inside the sculpture or beside it? Hiding them inside is more beautiful and makes us, on paper, the manufacturer of a new combined appliance whose approvals do not simply add up. Keeping them in a small separate hub the customer plugs together is defensible, honest, and lets the artwork stay an artwork. At twenty pieces this is a judgement, not a compliance programme, but it is yours rather than ours.

Three: which options are standard? The base build at €160 covers the room, a cable, and an opening with a sound system. Adding the WiiM at €150 covers the way most people play music at home. That is the one option I would argue should be fitted to every piece rather than offered — it is the difference between a frame that reacts to a room and a frame that a guest can play music to from their own phone in ten seconds.

Glossary

Terms used in this document, in plain language.

AirPlay

Apple's way of sending sound from an iPhone, iPad or Mac to a speaker over wifi. The button is already in Control Centre on every iPhone.

A2DP

The Bluetooth standard for streaming music, as opposed to phone calls. What a Bluetooth speaker uses.

Analysis path

Our term for sound coming in so the frame can make light. Very tolerant of poor quality.

Bluetooth SIG

The industry body that owns the Bluetooth name and standard. Charges a fee before any product may be sold as a Bluetooth device.

Dolby

A family of surround-sound formats. Decoding them requires a paid licence, which is why we arrange never to receive one.

Google Cast

Google's equivalent of AirPlay, used from Android phones and inside apps like Spotify and YouTube Music.

HDCP

The copy protection carried over an HDMI cable. It is what stops us reading sound from an Apple TV directly.

HDMI

The cable standard between a television and everything plugged into it. Carries picture and sound together.

Interface

Short for audio interface — a small box that turns analogue sound into something a computer can read. Ours has four inputs.

Line level

The standard signal strength that hi-fi equipment sends between boxes. Distinct from microphone level, which is roughly a hundred times weaker.

Listening path

Our term for sound going out to the customer's ears. Completely intolerant of faults.

MFi

Made for iPhone. Apple's licensing programme. Required to legally build an AirPlay receiver and to use the name.

Spotify Connect

The device list at the bottom of the Spotify app. Playing to a device there moves the music off the phone entirely.

WiiM Pro

An off-the-shelf €150 network music receiver that already holds the AirPlay, Google Cast and Spotify Connect licences.