How do you hide a picture in a sound?
Turn a picture into sound, or place it in your own recording, and see it appear in the spectrogram.
What you enter stays on your device while you use this tool. Opening a shared link sends the settings in its address to HowToolbox to prepare the page.
Spectrogram workspace
Alone, a picture becomes a sound of its own. Open a sound first to hide the picture in it.
Subtracts each frequency’s typical level, so hum and hiss fade and changing sounds stand out.
- Time step
- 10.7 ms
- Frequency step
- 23.4 Hz
Select a part to hear it alone: drag a rectangle, draw around a sound, or trace a gliding tone.
Speed, Normalise and Amplify apply when the selection plays (its own Play and the main Play over it).
Saves the selection, isolated, as a WAV file.
Choose Trace a curve and draw along a gliding sound to measure it: its pitch, glide and, for a passing source, its speed.
Sound above 18 kHz or below 40 Hz is brought into hearing when it plays. Slower or faster changes the rate: the call keeps its shape, and its pitch and length change together. Shift moves every frequency by a fixed amount and keeps the timing, like a heterodyne bat detector.
Alone, a picture becomes a sound of its own. Open a sound first to hide the picture in it.
Save it as WAV: MP3, AAC and other compressed formats smear or remove the picture.
Open a file or record a sound
Audio or video up to 500 MB, about 60 mins of sound at 48 kHz (less at higher sample rates). Nothing is uploaded. The sound is analysed on this device and never leaves it.
Turn a picture into sound, or place it in your own recording, and see it appear in the spectrogram.
Start with your own sound, the sample sound or a picture made into sound, and close a sound when you are done.
How to use
- Choose a picture: alone it becomes a sound of its own, shown in the spectrogram straight away. To hide it in a recording, open the recording first.
- Drag the picture to the time and pitch you want and drag a corner to resize it; its shape stays. Then press Mix in.
- Under Picture, set Dark parts loud for a dark drawing on white, White to black for its contrast, and its level against the sound; then download the track as WAV.
Where the rows of a five-row picture from 200 Hz to 8 kHz land
| Row from the top | Linear | Log | Mel |
|---|---|---|---|
| 1 | 7.22 kHz | 5.53 kHz | 6.23 kHz |
| 2 | 5.66 kHz | 2.65 kHz | 3.7 kHz |
| 3 | 4.1 kHz | 1.26 kHz | 2.1 kHz |
| 4 | 2.54 kHz | 605 Hz | 1.08 kHz |
| 5 | 980 Hz | 289 Hz | 429 Hz |
Row centres spread evenly on each scale. A placed picture has up to 256 rows.
How it works
A spectrogram draws sound as a picture: time across, pitch up, louder as brighter. This page runs it backwards. Each row of your picture becomes a steady tone at its own pitch and each column a moment of time, as loud as the pixel is bright; added together, the tones make a sound whose spectrogram is your picture.
Brightness becomes level in decibels: white is the loudest, black the chosen number of decibels quieter (White to black), and transparent parts are silent, so a logo keeps its outline. The rows are spread over the pitch range evenly on the frequency scale you place the picture on, linear in hertz, log in octaves or mel by ear, so it looks right on that scale and bends on another. Nothing goes above half the sample rate: 22 kHz for a 44.1 kHz track.
Mixed into a recording, the picture's loudness is set against the sound under it, and the sound in its box can be lowered so it reads clearly. Lossy formats such as MP3, AAC or a video upload throw away the faint parts it is made of; WAV keeps it. Thin lines blur a little, because a spectrogram trades sharpness in time against sharpness in pitch. Source: J. O. Smith, Spectral Audio Signal Processing (CCRMA, Stanford), additive synthesis.
Worked example
Where does the middle row of a picture placed from 200 Hz to 8 kHz land?
- Linear scale: halfway in hertz, (200 Hz + 8 kHz) ÷ 2 = 4.1 kHz.
- Log scale: halfway in octaves, √(200 Hz × 8 kHz) = 1.26 kHz.
- Mel scale: halfway by ear, which lies between the two: 2.1 kHz.
4.1 kHz on a linear scale, 1.26 kHz on a log scale and 2.1 kHz on a mel scale.
Sources
- Julius O. Smith III: Digital Audio Resampling Home Page (CCRMA)Used for: formula
- Julius O. Smith III: Mathematics of the DFT (CCRMA)Used for: formula
- Julius O. Smith III: Spectral Audio Signal Processing (CCRMA)Used for: formula
- Wikipedia: Bat detectorUsed for: definitions
Frequently asked questions
How can a sound contain a picture?
A sound is a mix of tones whose loudness changes over time, and a spectrogram plots exactly that. If the loudness of each tone follows the brightness of one row of a picture, the spectrogram shows the picture.
Why does the picture look bent in another app?
The rows are placed on the frequency scale you chose here. A viewer with another scale (many show linear by default) stretches the low or the high part. Place it on the scale the viewer uses.
Will the picture survive MP3 or a video upload?
Usually not well. Lossy formats remove quiet and very high sounds the ear barely notices, which is what a hidden picture is made of. Save and share the WAV file.
Why does it sound like noise?
A picture makes many tones at once, like a chord of hundreds of notes. Drawings and text with a few bright lines sound like chirps and whistles; photos sound like hiss. Keep the volume low.
Is my picture uploaded?
No. The picture is read, turned into sound and mixed on your device; nothing is sent or stored, and a copied link holds only the settings.





