Higgypop iconHiggypopSep 25, 2026 ~8 min source read

How to Use Spectrograms to Analyse EVP Recordings

A practical guide to reading spectral frequency displays, locating human voices in noisy recordings, and isolating or removing unwanted sounds with common audio editors.

Using Spectrograms To Analyse EVP Recordings

Share this story

Send the public story page.

Useful takeaways from this story.

Electronic Voice Phenomenon (EVP) are the mysterious sound of disembodied human-like voices of unknown origin that are heard through electronic devices.

The review and analysis of EVPs is normally carried out using free audio editing software like Audacity, or the more professional software Adobe Audition.

These applications generally show you the sound represented as a waveform by default.

# How to Use Spectrograms to Analyse EVP Recordings

What the spectrogram tells you

Recognising voice patterns makes it easier to find potential EVP material buried in noise. Speech often appears as clustered bands in the lower-to-mid frequency range with intermittent higher-frequency elements. With practice you will spot speech-like shapes amid other sounds.

Common noise types and how they look

  • White noise: appears as a uniform spread across the entire frequency range and time. It lacks distinct bands and looks like a constant smear.
  • High-frequency hiss: shows as lighter areas concentrated at the top of the spectrogram.
  • Low-frequency rumble: appears as energy concentrated toward the bottom.
  • Pure tones: show up as narrow, continuous bands at a single frequency, for example a 7 kHz tone or a 400 Hz tone used in demonstrations.

Seeing these shapes helps you choose edits that remove unwanted frequencies while preserving voice bands.

How to isolate and clean voices

Open your recording in an audio editor that offers a spectrogram view (free options like Audacity or paid options like Adobe Audition were mentioned). Use the spectral display to:

  • Identify the vertical placement of the voice and the noisy bands you want to remove.
  • Highlight the specific frequency range and the time interval where the noise occurs.
  • Delete or cut that selection, or reduce its volume to zero.

For narrow problems like clicks or short noises you can use targeted tools (the article mentions a 'spot healing brush' in editors that offer it) to blend edits so they sound less obvious.

  • A speech example concentrated under 500 Hz but with elements up to 6 kHz shows how voices occupy multiple bands.
  • Single-frequency tones (e.g. 7 kHz or 400 Hz) produce clear horizontal bands that you can remove without affecting other ranges.

Workflow tips

Repeated practice reading spectrogram patterns will make it faster to spot real speech versus random noise. The spectrogram does not prove a voice is paranormal, but it helps you isolate and clarify sounds for closer evaluation.

More context around this story.

Loading more related stories...

Keep reading in the app

Open the app view to save this story, compare related coverage, and continue from the same source.

Open in app