musicbmp

▶ Encoding

Audio samples are decoded to 24-bit integer, packed into BGR pixels, and laid out in a BMP image. Sample rate and bit depth are whispered into DIB header fields that no image viewer ever reads.

Each BGR pixel triplet stores one 24-bit audio sample. The image above is Beethoven's 9th — 96 kHz stereo, 3,072×1,680 pixels.

Visually, the encoded BMP appears as dense, colorful static — each pixel's color is a direct sample amplitude:

• Silence → pure black pixels (sample value 0)
• Loud passages → bright, saturated colors
• Stereo → alternating left/right channel columns create a subtle woven texture
• Frequency content → visible as rapid color transitions, with bass appearing as broad bands and treble as fine grain

Audio metadata (sample rate, bit depth, channel count) is embedded in DIB header fields invisible to image viewers.

▦ Masking

A visual mask becomes an audio filter. The mask image is normalized and multiplied sample-by-sample against the audio BMP — a photograph becomes a gate, a gradient becomes a fade curve, a texture becomes an amplitude modulator.

An encoded BMP, left to its own devices, is merely a canvas of static — beautiful in its chaos, but indifferent to the eye. Masking is the act of composing vision onto vibration. It is the moment a photograph leans in and whispers to a song: let me shape you. Suddenly, the raw pixel-noise of a waveform gains contour, texture, silhouette. A portrait becomes an envelope. A landscape becomes a crossfade. A hand-drawn scribble becomes a tremolo. The mask does not merely filter — it imprints. It grafts a visual narrative onto the fabric of sound, transforming a cold data grid into a collaborative artwork where every pixel pulls double duty: as both color and volume, both image and amplitude, both sight and sound.

◀ Decoding

The decoder reads the whispered metadata from DIB fields, unpacks each BGR triplet back into signed 24-bit audio, and writes a WAV file with the original sample rate and bit depth intact.