Soundliminal

What it is

Soundliminal generates a living, dreamlike soundscape, a perceptual shift rather than a product. It listens to the world around you through your phone's microphone, feels your movement through the gyroscope, reads the time of day and your surroundings; and from all of that, builds a sound world that has never existed before and will never exist again.

Every session is unrepeatable. The dream is always the same, and never the same.

What it's for

• Falling asleep in a city, without silence feeling like absence
• Working at the edge of focus, where the world softens but doesn't disappear
• Long commutes that feel like something more than transit
• Running at night when the world feels briefly enormous
• Moments when you want to feel like you're inside a film
• Listening to music and wanting it to mean something more

How to use it

Choose a dream from the library. Put on headphones. The illusion is designed for intimate listening and requires them. Let the dreamscape breathe with your surroundings for a few minutes before judging it; it takes time to weave itself around you.

Optionally add a song from the in-app catalog: it will play woven into the dream, and the dream will listen back, reacting to the song's energy, rhythm, and texture in real time.

Swipe right to leave. Drag up or down anywhere to shift the volume. Double-tap to locate any active satellites; their remaining time appears as a short code above the void.

The dreams

• Undertow, a slow current pulls the world underwater
• Neon, the city hums at a frequency you've never noticed
• Hollow, silence has a texture
• Static, the air between broadcast stations
• Void, a sensory absence that consumes whatever it catches and returns it, altered beyond recognition
• You, a dream generated fresh each day from your listening history — no two users ever hear the same one. Requires Apple Intelligence

The satellites

Four satellites can be sent into the void through optional in-app tips. Their presence keeps the app free for everyone else.

Relay satellites cross the screen as a train, a tight group of glowing dots traveling together at fixed speed, vanishing off one edge and reappearing on the other after a brief gap. Each tip adds 10 minutes; up to ten can be active simultaneously, their trains crossing in succession.

The GPS satellite arcs slowly right to left in a sine curve, periodically dropping a tracking signal downward. Each tip adds 20 minutes and stacks onto the remaining time.

The space telescope arrives from below the screen, traveling a curved Bézier arc to a parked position at the edge of the void. Once docked, it rotates imperceptibly in place, transmitting in morse-like bursts. Each dream open triggers a fresh arrival on a randomly chosen side. Each tip adds 30 minutes.

The exploration probe orbits in a slow Keplerian ellipse, faster near the event horizon, slower at apoapsis. Its three-dot body stays aligned along the instantaneous velocity tangent, periodically emitting a signal that curves down to the bottom of the screen. Each tip adds 30 minutes.

Double-tap the screen to locate the GPS satellite, telescope, and probe; their remaining time appears as a short code above the void, G20M for GPS, T1H15M for telescope, P2H30M for probe. Relay satellites travel as a group and have no locator.

Sound layers

Every dream is built from several concurrent layers that blend and shift in real time.

• Noise bed, a continuously filtered wash of broadband noise that forms the atmospheric foundation; its cutoff frequency, volume, and color shift with the dream's mood and the sounds around you
• Drone, a deep harmonic root tone that sustains beneath everything; when a song is playing, the drone reattunes to its key so both feel like they share the same gravity
• Crackle, a sparse layer of random impulse events, like static, vinyl surface noise, or distant sparks, that keeps the silence alive without ever repeating
• Phantom sounds, the most distinctive layer: timed sounds drawn from each dream's palette, appearing at random intervals and fading before you can be sure you heard them. Some are synthesized in real time, others are artist field recordings
• Dream echoes, fragments of what your microphone hears, captured silently, processed into something unrecognizable, and returned as part of the dream seconds later. What you said becomes part of the atmosphere. Nothing is stored or transmitted
• Song, any song from the in-app catalog can be layered into the dream; it passes through the dream's acoustic filter and becomes part of the soundscape rather than playing over it. The dream then listens back to the song through FFT analysis and reacts to its energy in real time
• Custom tones, each dream includes a curated set of synthesized tones unique to its character: underwater resonances, electrical hum, organic drones, field textures. They play as phantom events and never repeat exactly

What it can do

• Listens to the world, the microphone feeds ambient sound into the engine, so a noisy street, a quiet bedroom, or rain on a window all shape what you hear back
• Feels your movement, the accelerometer and gyroscope detect walking cadence, stillness, and orientation; the dreamscape pulses in rhythm with your stride when you move and settles when you stop
• Runs behind everything, Soundliminal plays as a background audio session, blending with whatever else is on your device. It can layer under a podcast, mix alongside another music app, or sit beneath your own playlist
• Reacts to songs in real time, live FFT decomposition of the playing song drives the visual satellite's orbit speed, color brightness, and comet emission rate. Louder passages produce more visual activity
• Reads time of day, the engine is aware of whether it's morning, evening, or deep night, and subtly shifts the dream's tone and density accordingly
• Adapts to your headphones, detects open, sealed, ANC, and spatial audio hardware and adjusts the mix so the illusion holds regardless of what you're wearing
• Simple to operate, swipe right to leave, drag up or down to shift volume, double-tap to locate satellites. No menus during playback. The experience is designed to disappear into itself
• Personalizable, each dream exposes independent parameters: noise level and cutoff, drone frequency and detune, crackle density, breath period, echo wet mix, phantom density and type distribution, song root frequency for drone reattunement
• Syncs settings across your devices, all parameters — volumes, EQ, motion sensitivity, constant noise mode — are automatically mirrored to your other devices via iCloud. Switch from phone to iPad and the dream picks up exactly where it left off
• Completely private, no accounts, no analytics, no network requests during a session. Everything happens on your device

Signal chain

Every layer is routed through a fixed graph before reaching your ears. The noise bed and song share a pre-mixer so they pass through the same low-pass filter together, making the song feel acoustically part of the dream rather than playing over it.

Reverb

Two independent reverb chains process different layer types.

Dream echoes use a cathedral reverb, the longest available decay:

The lunar factor means echoes are fuller and wetter under a full moon (lunarPhase ≈ 0.5), quieter and drier under a new moon (lunarPhase ≈ 0 or 1). The dream's echo character shifts with the moon without you noticing the cause.

Phantom sounds use a large hall reverb, applied after HRTF spatial positioning:

Each phantom is placed at a random 3D position before hitting the reverb, azimuth 0–360°, elevation ±31.5°, distance 0.8–5.0 m. The HRTF gives it direction; the reverb gives it room.

Spatial audio

Phantom sounds are placed in three-dimensional space using HRTF (Head-Related Transfer Function) rendering. HRTF is a mathematical model of how your outer ear, head, and shoulders filter sound differently depending on its direction. The result is that each phantom arrives from a specific point in the room, not just left or right, but in front of you, behind you, above or below, at a perceivable distance.

The effect is computed entirely in software. Any stereo earphone, including wired earbuds, gets full 3D spatial audio. AirPods add an improved HRTF model tuned to Apple silicon. AirPods Pro and AirPods Max additionally use head tracking: as you turn your head, the phantom positions stay anchored to fixed points around you, as if the sounds were objects in the room rather than effects in your ears.

Each phantom event is assigned a unique position before scheduling: azimuth drawn uniformly from 0–360°, elevation ±31.5°, distance 0.8–5.0 m. No two events share the same position. The spatial blend per event (how strongly the HRTF is applied) is drawn from 0.15–0.85, so some phantoms feel close and defined while others feel diffuse and directionless.

Drone synthesis

The drone is a seamlessly looping PCM buffer synthesized from scratch; no sample files. Six harmonic partials are summed with amplitude rolloff controlled by a tone parameter:

Amplitudes are normalized after summation so total loudness stays constant across tone values. Left and right channels use the same partials but different center frequencies, creating stereo shimmer:

The buffer length is chosen so the fundamental lands on a whole number of cycles, making the loop point inaudible:

Because k is an integer multiple, all harmonics also complete whole cycles within the same buffer length. The custom tone layer uses the same synthesis algorithm but is pinned to 528 Hz; real-time frequency selection is handled by a varispeed rate multiplier:

Dream echo processing

A rolling 6-second RAM buffer holds raw microphone audio; never written to disk. Two conditions trigger a capture:

The captured snippet (1.2–2.5 s) is peak-normalized, then returned to the dream in three waves at increasing delays:

Pitch shift is in cents (100¢ = one semitone). The direction is biased per dream; Undertow −0.8, Neon +0.5, Hollow +0.1, Static −0.3, You +0.05. So returned sounds trend toward each dream's tonal character. A speaker feedback gate mutes the echo bus entirely when the device is playing through its built-in speaker rather than headphones.

Live modulation

The engine updates every 50 ms (~20 Hz). Three forces act simultaneously on each tick:

Breath swell, a slow sine that gives the soundscape a living pulse:

Rhythm duck, where detected footfall pulses compress layers in time with walking cadence:

Ambient and motion:

The solar factor (0 at midnight, 1 at midday) opens the filter during the day and closes it at night; the dream's texture darkens as the light does. Song energy compresses phantom scheduling intervals:

Louder song passages produce more phantoms, as if the music were shaking them loose from the dream.

Per-dream parameters

All six dreams run through the same engine. Their character is entirely defined by these signal parameters:

• Undertow, noise 0.50 @ 260 Hz · drone 55 Hz, detune 0.8, level 0.35 · no crackle · breath 9 s · echo wet 80%, pitch bias −0.8 · phantom density 0.35
• Neon, noise 0.22 @ 1600 Hz · drone 110 Hz, detune 2.0, level 0.40 · crackle 3 · breath 6 s · echo wet 55%, pitch bias +0.5 · phantom density 0.70
• Hollow, noise 0.45 @ 900 Hz · drone 220 Hz, detune 0.6, level 0.18 · no crackle · breath 11 s · echo wet 85%, pitch bias +0.1 · phantom density 0.50
• Static, noise 0.30 @ 6500 Hz · no drone · crackle 25 · breath 4.5 s · echo wet 35%, pitch bias −0.3 · phantom density 0.80 · echo level override 0.60
• Void, no noise bed · no drone · no crackle · echo wet 92%, cooldown 0.6 s, capture 0.3–6.0 s, up to 6 returns · phantom density tracks ambient silence · all 24 synthesized phantom types · cathedral reverb 90% wet · spectral analysis of each captured echo biases which phantom type fires next
• You, all parameters AI-generated daily from your 7-day listening history · phantom pool drawn from all four dream palettes · breath 8–20 s · echo wet 75% · requires Apple Intelligence

The noise cutoff defines each dream's fundamental color: Undertow is almost entirely sub-bass murmur; Static is nearly all treble hiss. Crackle density controls how alive the silence feels; Hollow and Undertow are clean, Static is almost tactile. Detune on the drone controls shimmer: Neon's 2.0 Hz offset creates a distinctly urban beating quality; Undertow's 0.8 Hz is barely perceptible, just enough to feel alive.

Environmental noise handling

A 2048-sample FFT runs on the microphone buffer once per second. If a constant environmental tone is detected, such as fluorescent hum, air conditioning, or a mechanical fan, the engine chooses one of three responses:

• Sidestep, a parametric notch at the detected frequency: −14 dB, bandwidth 0.7 octaves. The dream continues unimpeded while the environmental tone is quieted beneath it
• Swallow, a counter-tone at the exact frequency injected into the mix in 3–8 s bursts, followed by 15–60 s silence. In sealed headphones this creates a partial null at the intruding frequency
• Absorb, a broad +3 dB boost (bandwidth 1.5 octaves) around the noise frequency, giving the dream presence in the same band and perceptually masking the intrusion by raising the dream above it

The math

The probe's orbit is computed frame-by-frame from Kepler's equation, solved by Newton–Raphson iteration:

where M is mean anomaly, E is eccentric anomaly, ν is true anomaly, e ∈ [0.05, 0.35] is orbital eccentricity (drawn fresh each orbit), and a is semi-major axis. Velocity direction is computed as a forward finite difference (Δν = 0.006 rad), aligning the three-dot probe body tangent to its instantaneous trajectory.

The GPS satellite follows a sine arc: y = baseY − h·sin(π·progress), where h (arc height) is seeded per transit using a linear congruential generator, seed = (cycleN × 1664525 + 1013904223) mod 2³¹, giving stable, reproducible randomness without any stored state. The space telescope instead travels a quadratic Bézier from below the screen to a fixed parking position at the edge of the void; its body orientation during transit is the derivative of the curve, dtx = 2(1−t)(cx−sx) + 2t(px−cx), dty likewise, so it always faces the direction of travel. Once parked, it rotates at a constant ~0.46°/s, barely perceptible.

Signal dots follow a quadratic Bézier-like curve: x(t) = (1−p)²·x₀ + (1−(1−p)²)·xc, y(t) = y₀ + p²·(yc − y₀), where p ∈ [0,1] is normalized time, (x₀, y₀) is the emission position, and (xc, yc) is the bottom center of the screen.

The tone layer draws from the nine solfeggio frequencies, a set of intervals whose digit sums reduce exclusively to 3, 6, or 9, the pattern Tesla described as the key to the universe:

Drone frequencies and phantom tones are tuned to octave multiples of these values, so every harmonic produced by the engine sits on a node of the same numerical lattice. The root drone, when not overridden by a song's key, defaults to 174 Hz (or its sub-octave at 87 Hz), the lowest solfeggio, associated with grounding and physical sensation.

The connection to Chladni figures is not metaphorical. Ernst Chladni showed in 1787 that a plate vibrating at a resonant frequency organizes sand into standing-wave nodal lines, the geometry of the sound made visible. The governing equation is the biharmonic wave equation for thin plate flexure:

where w is the plate displacement, D the flexural rigidity, E Young's modulus, h plate thickness, ρ density, and ν Poisson's ratio. The eigenfrequencies of this system, meaning the values at which the plate rings, depend on its geometry and boundary conditions. The solfeggio frequencies sit near the eigenfrequencies of certain rectangular and circular plates of natural materials, which is why they were rediscovered empirically across many cultures long before the equation was written.

In Soundliminal the orbiting particles around the black hole are a temporal Chladni figure: the ring of matter that accumulates at the photon sphere is where the gravitational wave equation's solution has a nodal surface, matter that neither falls in nor escapes, tracing the geometry of the field in the same way sand traces the geometry of sound.

Disclaimer

Soundliminal is an artistic and creative audio experience. It is not a medical device and is not intended to diagnose, treat, cure, or prevent any physical or mental health condition. References to solfeggio frequencies, resonance, and their associated names reflect historical and cultural tradition, not medical or therapeutic claims.

Listen at a safe volume. Prolonged exposure to audio at high levels can cause permanent hearing damage. The app is designed for intimate headphone listening. If others nearby can hear your headphones, the volume is too loud.

Do not use Soundliminal while driving, cycling, operating machinery, or in any situation that requires your full attention. The experience is intentionally immersive.

Support

Soundliminal is built and maintained by one person. If something doesn't feel right, write to andresrobert@icloud.com.

Summary

Soundliminal does not collect, store, or transmit any personal data. There are no accounts, no analytics, no tracking, and no network requests made during a session. Everything the app does happens entirely on your device.

Microphone

The app requests microphone access to listen to your ambient environment. This audio is processed in real time within the app to shape the dreamscape; a loud room may shift the noise bed, and quiet moments allow subtler sounds to emerge.

The Dream Echoes feature captures short fragments of ambient audio in RAM, processes them into unrecognizable textures, and plays them back as part of the sound experience within the same session. These fragments are never written to disk, never transmitted, and are automatically discarded when the session ends.

Microphone access is optional. The app functions without it, but the environment-reactive features will not be available.

Motion sensors

The accelerometer and gyroscope are used to detect movement, walking cadence, stillness, and orientation. This data is read locally and used only to modulate the audio in real time. It is never stored or transmitted.

Location

The app can optionally request location access to read your GPS coordinates. This is used only to compute the current solar elevation angle (which shapes the dream's brightness over the course of the day) and the lunar phase and compass bearing (which influence reverb depth and stereo pan bias). No coordinates are stored, logged, or transmitted. Location access is off by default and can be toggled in Settings.

When location is disabled, the same calculations fall back to the device clock alone: time of day stands in for solar angle, and lunar phase is estimated from the calendar date.

What is not collected

• No personal information of any kind
• No location data stored or transmitted — GPS coordinates are read locally and discarded immediately
• No usage analytics or crash reporting sent to any server
• No advertising identifiers
• No third-party SDKs that collect data
• No account creation or login
• No audio recordings stored or transmitted

In-app purchases

Optional satellite tips are processed entirely through Apple's in-app purchase system. Soundliminal does not handle or store any payment information. Apple's privacy policy governs those transactions.

Contact

Questions about privacy can be directed to andresrobert@icloud.com.

Last updated: July 2026