Audio Spectrogram Message Hider

Audio spectrogram steganography is the practice of hiding messages inside audio files by encoding them as specific frequency patterns. When the audio is viewed as a spectrogram (a visual representation of frequencies over time), the hidden message appears as horizontal bars at different frequency positions. Unlike traditional audio steganography (LSB encoding), spectrogram messages are visible to anyone who looks at the spectrogram but are not obvious during regular audio playback.

Each printable ASCII character (codes 32-126) is mapped to a unique frequency between 300 Hz and 1060 Hz. The mapping is linear: the space character maps to 300 Hz, exclamation mark to 308 Hz, and so on up to tilde (~) at 1060 Hz. The 8 Hz spacing between characters ensures clear separation in the spectrogram, making each character's frequency visually distinct as a separate horizontal bar.

The decoder is designed to extract messages encoded with our specific frequency encoding scheme (300-1060 Hz range, 8 Hz spacing, with 2000 Hz start marker and 2500 Hz end marker). Random audio files will not contain these specific frequency patterns and will likely return no message or gibberish. The decoder provides a confidence score for each detected tone to help assess reliability.

Partially. Our WAV-based encoding uses frequencies in the audible range, and moderate MP3 compression (192 kbps and above) may preserve enough of the frequency content for successful decoding. However, heavy compression, downsampling, or aggressive noise reduction can distort the frequency patterns and reduce decoding accuracy. For best results, use WAV format for both encoding and decoding.

Each character requires approximately 0.5 seconds (0.4 seconds of tone + 0.1 seconds of padding). A message of 20 characters produces a 10-second audio file. There is no hard limit on message length, but very long messages produce correspondingly long audio files. Short messages (under 100 characters) work best for clear spectrogram readability.

The encoding uses frequencies in the 300-2500 Hz range, which is well within the human hearing range (20-20000 Hz). Each character is represented by a 0.4-second sine wave tone, so the audio will sound like a sequence of beeps or whistles at different pitches. This is expected behavior — the message is meant to be visible in the spectrogram, not inaudible.

The Goertzel algorithm is a digital signal processing technique that efficiently detects the presence of specific frequencies in an audio signal. Unlike the FFT (Fast Fourier Transform) which computes the entire frequency spectrum, Goertzel can target specific frequencies of interest, making it faster and more accurate for detecting the discrete character frequencies used in our encoding scheme.

While this tool can encode and decode messages in audio spectrograms, it is designed primarily for educational and demonstrational purposes. The encoding pattern is deterministic and easily recognizable once the method is known. For serious security applications, consider using established encryption tools combined with more sophisticated steganography techniques.

Yes, completely. All processing — audio generation, spectrogram rendering, and frequency detection — happens entirely in your browser using JavaScript, the Web Audio API, and Canvas API. Your messages, audio files, and generated spectrograms are never sent to any server. No account, signup, or internet connection is required beyond loading the page.