Parallax Cipher

The visual layer: parallax as the cipher

The "surface" is an arrangement of bodies and innocuous objects in a real space — a plaza, a queue, a market, a march. The reconstruction mechanism is **occlusion and alignment**, both of which change with sightline. Four concrete encoding methods:

**Occlusion-gated tokens.** Each participant carries something semantically neutral — a colored bag, a hat, a banal sign. Scattered through the crowd these read as random noise from the civilian angle. From the keyed vantage, depth causes most tokens to vanish behind bodies, and only a specific visible subset remains — and *that* subset is the message. It's a physical moiré: the right alignment filters the field down to signal.

**Collinearity encoding.** From the correct point, depth collapses and certain people *line up* — heads forming a row, a color sequence reading cleanly left-to-right because foreground and background flatten into apparent adjacency. From any other point the depth spreads them out and the order scrambles. The "image" is an emergent alignment, exactly as in a hologram.

**Depth-plane layering.** This is where your "complicated message in layered parts" comes in. Like a multi-plane hologram, you assign messages to different depths. The reader at vantage X (close, oblique) resolves plane 1; the reader at vantage Y (far, frontal) resolves plane 2. One crowd carries several distinct messages, each addressed to a different standing-point — a different recipient at a different bench reads a different thing from the identical scene.

**Anamorphosis.** The Holbein-skull / street-chalk-art trick: fragments held by separate people assemble into legible form only from one steep oblique angle, and are pure distortion otherwise.

The deniability is structural, not added on. From any non-keyed angle the scene *is* an ordinary crowd. A photograph taken from the wrong place isn't weak evidence — it's exculpatory, because the message has no existence there to capture.

## The audio layer: combination as meaning

The acoustic analog of viewing angle is **listener position in a sound field**, plus **combinatorial co-occurrence**. Several mechanisms, which stack:

**Spatial sweet-spot resolution.** Multiple ordinary sources (a radio here, a phone there, a vendor's call) at different locations sum differently depending on where you stand. At one point the layers balance into something legible — a phrase, a dominant track, a rhythm; a step away it's babble. This is a literal acoustic hologram, the same principle as wave-field-synthesis sweet spots.

**Polyphonic / chord meaning.** No single sound is suspicious — a cough, a ringtone, a specific song on a specific station. The meaning is in the *set and the ordering*, like a chord versus its notes. "Which three songs, in which sequence" or "which stations are tuned simultaneously" is the signal; the components are deniable individually. This is your "combination of sounds includes additional meaning."

**Timing as carrier.** The gaps between innocuous events — when each radio switches on, the pauses between calls — encode bits, Morse hidden inside ordinary activity. The data is in the silence.

**Channel diversity.** Phones and radios let the "voices" be geographically scattered. A person repeating a banal phrase on a call, a radio on a particular track, a text sent at a particular second — the *intersection across channels* is the message, while each channel alone is nothing.

## Cross-modal layering: the elegant part

The two channels shouldn't just coexist — they should **index each other**, which is what makes the combined system far stronger than either alone:

- **Time-gating.** The audio defines *when* the visual alignment is live. The crowd-message only "exists" during a particular song or tone. Even standing in the right place, you read nothing in the wrong acoustic window.

- **Disambiguation.** The visual provides what you called "a set with potential for communication within it" — a frame, a space of readings. The audio is the pointer that selects which reading is active. The crowd is the codebook page; the sound is the index. One static arrangement can mean many things across an afternoon depending on the audio key.

- **Mutual error-check.** A message counts as valid only if *both* modalities resolve consistently. This kills accidental false positives and makes spoofing require defeating two unlike systems at once.

- **Bandwidth split.** Vision carries the slow structural part (topic, frame); audio carries the fast variable part (the specific instruction).

## Procedure

The decode key is the **vantage plus the window** — where to stand and when to listen. So the operational core is distributing that key without exposing it: an innocuous landmark reference ("from the third bench, while the second song plays") *is* the cipher key, and only key-holders can read. Two regimes:

- **Choreographed** — participants know they're forming a message (rehearsed like stadium card-stunts/tifo, but angle-gated). Higher fidelity, but more knowing participants means a larger conspiracy and weaker deniability.

- **Unwitting** — only the arranger and the reader know meaning; participants get only placement and prop instructions ("stand here, wear blue"). Far more deniable and secure, lower bandwidth, harder to steer.

## Honest limits

It's worth being clear about the threat model so the design isn't fantasy. This resists a **single-vantage** or **single-photograph** observer — and that's exactly the property that gives legal/evidentiary deniability. It does *not* resist a comprehensive adversary: many cameras at many angles, or an observer free to move, can image the whole "hologram" and recover every plane, just as fully imaging a real hologram reveals everything. Bandwidth is low — this is signaling (go/no-go, timing, identity confirmation, short codes), not data transfer. And it requires a coordination channel to set up, which is itself the softest point of attack — the same key-distribution problem every cipher has.

If you're building this into fiction or a framework, the richest tension lives right there: the system's strength (it doesn't exist from the wrong angle) and its fatal weakness (it fully exists to anyone who can occupy every angle) are the same fact. Happy to push on any one layer — the depth-plane addressing, the acoustic sweet-spot math, or the unwitting-participant logistics — if you want to develop it further.