Updated 2026-11-15

Spectrogram Decoder & Image to Sound

This spectrogram decoder shows the pictures people hide inside sound. Load a clip, zoom into the frequency band and raise the contrast until the image appears, then save it as a PNG. Switch to encode mode to turn your own picture or text into sound. Nothing is uploaded.

  • Free, no signup
  • Decode and encode
  • PNG and WAV export
Spectrogram decoder — Decode audio
Runs locally · 0 uploads

Secret message (typed text)

–:–– / –:––
Frequency window0 Hz – 11 kHz
Frequency rows
FFT size21.5 Hz
0 dB
-89…-20 dB
Colour mapMagma
PNG size
Space play0 fit

Built in

Two tools in one window

Read pictures out of audio, or write them in. Both directions run on your device, with the same frequency maths.

  • 0 uploadsAudio and pictures stay in your browser
  • 100 MBAudio files up to 50 MB free, 500 MB with Pro
  • 20 Hz – 20 kHzEncode into any band you choose
  • PNG · WAVDecoded PNG up to 1920 px wide free, 3000 px with Pro
Spectrogram decoder view of a photo hidden in sound, in the ice colour map

Decode: audio → picture

A spectrogram decoder tuned for pictures, not music: pick the frequency window, set brightness and contrast in dB, invert, then drag a box to zoom.
  • FFT 1024–4096
  • Linear
  • Log
  • Invert
Colourful long-exposure light trails on a black background

Encode: image to sound

Turn a picture or typed text into a WAV whose spectrogram shows it. Choose the band, length and sound, or hide it inside a song.
  • 2–30 s
  • 128–400 rows
  • WAV 16-bit

Examples

Hidden images in audio, found with this decoder

Every picture below was encoded with the tool's image to sound mode and decoded again with the spectrogram decoder. Play the clip, then open it in the decoder with the same settings.

Full-range spectrogram of a song with a faint hidden drawing near the top
Spectrogram decoder zoomed to 13–19 kHz showing a hidden cat drawing in the song
Full range13–19 kHz, contrast up

A cat hidden in a pop song

The full view shows a faint shape above the music. Narrow the window to 13–19 kHz and raise the contrast, and the drawing is plain to see. It sits 24 dB under the song, so you barely hear it.

  • 13.5–18.5 kHz
  • −24 dB
  • MP3 320 kbps
  • Ice
Spectrogram decoder showing the words YOU FOUND IT hidden in an audio clip

Typed text, decoded

“YOU FOUND IT” written with 140 tones between 2 and 9 kHz over 8 seconds. Magma, linear rows.

  • 2–9 kHz
  • 140 rows
  • Linear
  • Magma
Image spectrogram of a radio telescope photo encoded as sound and decoded again

A photo, drawn with 300 tones

A radio-telescope photo as 300 sine waves from 1 to 16 kHz. Grayscale reads a photo best.

  • 1–16 kHz
  • 300 rows
  • FFT 2048
  • Grayscale
Grainy spectrogram picture made with the noise method of the image to sound encoder

Same photo, noise method

Inverse STFT with random phase fills the gaps between rows, but the picture turns grainy.

  • 1–16 kHz
  • iSTFT
  • FFT 2048
  • Grayscale
Hidden cat drawing shown inverted as dark lines on a light spectrogram

Inverted: ink on paper

Invert turns loud into dark. Line drawings and text often read better this way.

  • 13–19 kHz
  • Invert
  • Grayscale
Text in a spectrogram distorted by viewing it on a log frequency scale

Wrong row spacing

The same text viewed with log rows: the top squeezes and the bottom stretches. Match the scale the picture was made with.

  • 20 Hz – 11 kHz
  • Log
  • Magma

Want the usual view of a song, with notes and dB readouts? Use the online spectrogram. For a clean outline of the loudness instead, make a waveform image.

How it works

Find a hidden image in audio in three steps

  1. Load the audio

    Pick a demo clip or drop your file. A lossless WAV or FLAC keeps the most detail.

  2. Zoom into the band

    Drag a box over anything that looks drawn, or press Find picture. The upper band is the usual hiding place.

  3. Tune and save

    Raise the contrast, try invert or grayscale, then download the decoded picture as a PNG.

Settings

What each image spectrogram control does

A spectrogram turns sound into an image: time runs left to right, frequency bottom to top, loudness becomes brightness. These controls decide which part of that image you see.

The hidden cat filling the plot after narrowing the frequency window to its band

Frequency window

Show only the band the picture lives in, so it fills the plot instead of a thin strip.
  • 0–22 kHz
  • 12 kHz +
  • Find picture

Brightness and contrast

They move and narrow the dB window. A picture hidden 25 dB under the music needs both turned up.
  • ±30 dB
  • 110 → 20 dB

Linear or log rows

Most hidden images are drawn on linear rows. Log suits music, and pictures made for it.

FFT size

4096 resolves fine rows, 1024 keeps quick changes sharp. 2048 is a good start.
  • 1024 · 2048 · 4096

Invert and colour maps

Grayscale suits photos, invert suits text and line art. Magma and Ice help faint shapes stand out.
The same photo encoded as tones on the left and as noise on the right

Image to sound: tones or noise

Tones give a sharp picture with fine lines. Noise fills the image evenly and sounds softer.
  • Tones (sharp)
  • Noise (soft)

Hide it in a track

Mix the picture into the demo song or your own track at a level you pick, then export a WAV.
  • −40…−6 dB
  • WAV

Use cases

Who decodes pictures in sound

From puzzle hunts to studio easter eggs, these are the usual reasons to open a spectrogram decoder.

Magnifying glass and compass on an old map, like a puzzle hunt

Puzzle hunts and ARGs

Alternate-reality games often hide a clue in an audio file. Drop it in, check the upper band and screenshot what you find.
Music producer working at night in a blue-lit home studio

Easter eggs in music

Producers sign tracks with logos, faces and words in the spectrum. Load the song and look for shapes that don't move like music.
Close-up of a vintage shortwave radio tuning dial and knob

Radio and signal hobby

Recordings from radio hobbyists sometimes carry drawings or text in the waterfall. A narrow window and log rows help.
Person wearing headphones in a dark room, seen from behind

Send a secret message

Type a name or a date, turn it into sound and send the WAV. Your friend opens it here and reads the message.
Colourful patch cables plugged into a modular synthesizer

Sound design

A picture played as sound is a strange, glassy texture. Encode shapes and gradients and use the WAV as raw material.
Instant photo prints on envelopes on a wooden table

Teach how sound works

Encode a photo, play it, decode it. It shows students that every row is a frequency and every pixel a loudness.

Photos: Unsplash (Unsplash License). Demo clips, the cat drawing and all spectrogram images: made with this tool.

Guide

How a hidden image in audio works — and where to look

A spectrogram slices sound into short frames and measures how loud every frequency is in each one. Plot the result and you get a picture. Turn that around and you can paint a picture first: make each row of pixels a frequency, play that frequency as loud as the pixel is bright, and the spectrogram draws your image back. That is all an image to sound encoder does.

Artists have used the trick for decades. Aphex Twin hid a face in his 1999 Windowlicker single, Nine Inch Nails tucked images into the Year Zero campaign in 2007, and Mick Gordon's Doom soundtrack from 2016 carries a pentagram and the number 666. Puzzle makers do the same with clues, which is why a spectrogram decoder is a standard tool in ARG communities.

When nothing shows up at first, work through the usual suspects. Start with the full view and the 12 kHz + preset. Raise the contrast until the background turns dark. Try log rows if the picture looks squashed at the top, and invert if it is text. If the file is an MP3 and the top of the image is missing, the encoder has already thrown that band away; ask for the WAV.

New to spectrograms? Read how to read a spectrogram first.

Where pictures usually hide

BandWhat you hearSurvives MP3?
15–20 kHzFaint whistle, inaudible to many adultsOften cut at 128–192 kbps
5–15 kHzHiss or screech over the musicYes, a little blurred
Below 2 kHzClearly audible, clashes with the musicYes

Making your own

Simple, high-contrast art encodes best: thick white lines on black, or bold text. Keep the picture between 1 and 15 kHz if it will be shared as MP3, use 256 rows or more for photos, and give it at least 8 seconds so the picture isn't squeezed.

FAQ

Spectrogram decoder questions

A spectrogram decoder draws a sound as a picture of frequency over time and gives you the controls to read what is drawn in it: a frequency window, brightness, contrast, invert and colour maps. If someone encoded an image into the audio, the decoder makes it visible.

Load the file, look at the full view, then try the 12 kHz + preset or Find picture. When you see a shape, drag a box around it to zoom and raise the contrast. Text often reads better inverted. Use the original WAV or FLAC if you can.

Yes. Switch to Image to sound, pick a picture or type text, choose the length and frequency band, and press the button. Each pixel row becomes a tone. Download the WAV, or hide the picture inside the demo song or your own track first.

MP3 encoders drop high frequencies to save space, often everything above about 16 kHz at 128–192 kbps, and they blur quiet detail. Keep hidden images below 15 kHz for MP3, or share the WAV.

Use the scale the picture was made with. Almost every hidden image is drawn on linear rows, which is also this encoder's default. On log rows such a picture looks squeezed at the top.

It is free with no signup and no watermark, with 10 PNG or WAV downloads a day; Pro lifts that limit and adds the 3000 px export. Decoding, encoding and the PNG and WAV exports all run in your browser, so your audio and pictures never leave your device.

Find the picture in your audio

Drop a file into the spectrogram decoder, or turn your own image into sound.

Free · no signup · no watermark · processed in your browser