SpectraTrace

Seven instruments for the investigation.

SpectraTrace is an iPhone app for ghost hunts and paranormal investigations. It turns the phone's own magnetometer, microphone, camera, touch screen and GPS into instruments that keep measuring for the whole session.

Every event is stamped with its time, its source and its size, so what you review afterwards is the reading, not the memory. Each instrument's guide cites the independent studies behind what it measures.

M1 Simulated signal for this page
Δ0.0 µT
Base48 µT
Noise0.21
Events0
Samples0
Get it on the App Store €2.99One purchase. No subscription, no ads.
SpectraTrace home screen on the chart-recorder design: the Hotspot Map card at top, then the six other instruments listed below it
The instrument list, redesigned

Why it matters while the session is running

A session is short, dark and full of distractions. The instrument's job is to keep measuring when you can't, and to keep a record you can trust when the group disagrees about what happened.

  1. Before: learn the room and the ground

    M1 learns the resting magnetic field in about two seconds and keeps adapting to slow drift. Check the Hotspot Map for what is actually nearby — power lines, historic sites, water, geology — so a reading has a baseline and a place to start explaining it.

    Base 48 µT · Noise 0.21

  2. During: every input, one timeline

    Field, thermometer, microphone and touch each raise their own events. REM turns them into spoken prompts, Spectrum turns the magnetic field into sound, and Frequencies shows the sound and transcribes speech live. Nothing is silently merged: each event names its source.

    05:30:17 House · Field · Δ 8.6 µT

  3. After: review the numbers

    Sessions, words, audio and video stay on the phone. Export M1 and Frequencies as CSV, replay the recordings, and compare a quiet control period with the moment everyone remembers.

    frequencies_2026-09-15.csv

Seven instruments, one rack

Each one measures something real with a specific part of the phone. Under every instrument: what it uses, and the independent work that tells you what such a measurement can and cannot show.

New in 1.1

Hotspot Map

Theory-weighted area index

A live heat map of the surroundings, built from OpenStreetMap and USGS data: power lines and transmitters, historic and burial sites, rivers and water, rock and faults, combined into a 0–100 index by four popular theories. The index follows your GPS as you walk, and you set the theory weights yourself.

Measures with
Public OpenStreetMap and USGS data, matched to a rounded grid point near you rather than your exact position
Backed by
The index is an area score built from the theories you choose to weight, not a paranormal measurement — the guide is explicit that it names nearby features, not causes.
Hotspot Map screen showing a scanning radius over a street map and the theory-weighted index below it
M1 screen: readouts row, the field chart on ruled paper, and the input chips for thermometer, microphone and touch

M1

Magnetic field meter

Three-axis magnetometer readings 50 times a second, combined into field strength and compared with a learned baseline. Live chart on a log scale, adjustable sensitivity, event detection with hysteresis so one disturbance isn't counted five times. Every sample can be exported.

Measures with
Core Motion three-axis magnetic field
Backed by
Odenwald 2022, Space Weather and Hellesund 2019, Physics Education: smartphone magnetometers resolve real field variation under stationary conditions; bias and noise are device-dependent.
REM screen with a printed word on ruled paper and the event log below it

REM

Sensor-triggered word prompts

When any enabled input crosses its threshold, REM picks a random word from the list, shows it and speaks it. The word is a prompt for the group's observation, never a decoding. Each accepted word is saved with the time, the source and the size of the event that triggered it.

Measures with
The same magnetic detector as M1, plus the optional inputs
Backed by
Nees & Phillips 2015, Applied Cognitive Psychology: expectation shapes what listeners hear in ambiguous material — which is why REM labels its words as random stimuli and keeps the trigger value.
Spectrum screen with the ten-band analyser printed on ruled paper

Spectrum

The magnetic field as sound

A continuous noise bed whose brightness and level follow the magnetic field, so you can hear a change without looking at the screen. The analyser shows what is playing and doubles as a ten-band equalizer: tap an octave to cut it from the noise and clean the sound. The noise is generated on the phone; nothing is received from radio, and nothing is spoken — word prompts belong to REM.

Measures with
M1's field pipeline, mapped to sound (sonification)
Backed by
Nees & Phillips 2015: broadband noise is exactly the ambiguous sound in which listeners who expect a message report voices. The guide says so, and the noise carries no words.
Frequencies screen with the live spectrum, the voice band marked, and the level readouts below it

Frequencies

Live audio spectrum with speech-to-text

The microphone from 20 Hz to 20 kHz, with the voice band marked and an event rule for energy that rises outside it. Turn on Speech to Text and spoken words appear in the transcript as they are said, so the session log and what was said in the room line up.

Measures with
Accelerate DFT with a Hann window, Apple Speech
Backed by
Standard signal-processing practice for windowed spectra; the guide explains why dB values here are relative, not calibrated sound levels.
Spectra Cam screen: camera view of a raised hand with the SLS hand-pose skeleton drawn over it

Spectra Cam

Camera, filters, body and hand pose

Record video with normal and mono views, an SLS overlay that draws the body and hand poses the software detects, and a live magnetic readout. The overlay is recorded into the video; the guide is explicit that this is pose detection, not thermal imaging.

Measures with
Vision body-pose and hand-pose detection
Backed by
Apple's own accuracy notes: the detector can miss people and return inaccurate poses, so an overlay is a classification to check against the picture, not a presence.
Audio RePlayer screen listing recordings, with search and export controls above the list

Audio RePlayer

Recording library and playback

Every recording made in M1, REM, Spectrum and Frequencies, kept as the original file. Play it back, share it, delete it. The app never applies a hidden voice detector or interpretation to what you recorded.

Measures with
The microphone, uncompressed PCM in a .caf file at its native sample rate
Backed by
Nees & Phillips 2015: listen once without a suggested transcript, write down what you hear, and have someone else do the same independently.

The studies behind the instruments

SpectraTrace's guides cite peer-reviewed work and framework documentation, and say plainly what each source supports. That boundary is the product: measurements you can defend, and no claims you can't.

Can Smartphones Detect Geomagnetic Storms?

Odenwald, S. F. — Space Weather 20, e2020SW002669 (2022)

phone magnetometers register real magnetic-field variation when the phone is stationary; sensitivity, noise and baseline behaviour differ between models, and charging disturbs readings.

that any change has a paranormal cause, or that a given threshold identifies one.

doi.org/10.1029/2020SW002669

Measuring Earth's Magnetic Field Using a Smartphone Magnetometer

Hellesund, S. — Physics Education 54, 025014 (2019)

the measurement itself, and the orientation-dependent internal bias and magnetic contamination that M1's baseline and recalibration are designed around.

that a phone can locate the source of a field change.

doi.org/10.1088/1361-6552/aafa6e

Auditory Pareidolia: Effects of Contextual Priming on Perceptions of Purportedly Paranormal and Ambiguous Auditory Stimuli

Nees, M. A. & Phillips, C. — Applied Cognitive Psychology 29, 129–134 (2015)

that paranormal framing increases how often listeners report voices in ambiguous audio, with low agreement about content — the reason REM labels its words as random stimuli, Spectrum’s guide warns about voices heard in noise, and the recorder preserves originals.

anything about this app specifically; it is a study of perception.

doi.org/10.1002/acp.3068

Framework documentation

Apple Developer — Core Motion, Accelerate, Speech, Vision, Core Image, AVFoundation; OpenStreetMap and USGS — public map and earthquake data

the units, interfaces and known limits of every sensor and detector the app uses, and the coverage and licence of the public data behind the Hotspot Map. Each guide links the exact page.

the app's own thresholds and event rules, which the guides describe as implementation choices.

developer.apple.com — CMMagneticField

Details

€2.99One purchase. No subscription, no ads, no account.
5 languagesEnglish, Italiano, Français, Deutsch, Español — word lists and voice follow the app language.
OfflineEverything runs on the phone. The Hotspot Map fetches each new area once, then keeps it. Privacy policy.
iPhoneNeeds a magnetometer (every iPhone since the 3GS). iOS 17 or later.
+ thermometerOptional Bluetooth LE thermometer: Govee, Xiaomi ATC/pvvx, SwitchBot or any standard GATT thermometer.
CSVExport M1 and Frequencies sessions with every sample, baseline, threshold and event flag.
Haptics offOff by default in M1 and REM, so the vibration motor doesn't disturb the magnetometer.
Public map dataHotspot Map uses OpenStreetMap and USGS, matched to a rounded grid point rather than your exact position.