Synth Poems Details

Synth Poems are short (1 minute) generative audiovisual pieces. At the time of minting, a random value is stored in an ERC721 token. This value serves as input to an unsupervised generative model that produces sound and animation. The code(written in C, ISO C99) that describes this model is stored directly on the Ethereum blockchain. Our server, as convenience, runs the code and caches the output as an MP4 video. However, anyone may read the publically available code, compile it and execute to reconstruct the output. For even more detail about this method of on-chain storage, see the About page.

At the musical level, the generative algorithm plays with feedback to produce an evolving but balanced set of timbres and pitches. Across pieces, a carefully selected set of rules govern the variation of tempos, time signatures, (in)harmonicity resulting in individual pieces that are unique, while retaining a common theme.

At the audio synthesis level, the code model makes use of Frequency Modulation(FM) synthesis. FM is an encoding method originally used for analog signals in radio broadcasting, and later discovered to be useful for audio synthesis allowing a simple way to generate rich harmonic timbres. Don Buchla made use of FM in his modular synthesizers of the 60s/70s, characterizing the "west coast" synth sound.

At the visual level, the code model simulates the action of an oscilloscope in XY mode. The left and right stereo audio channels act as inputs to the XY inputs of a (virtual) oscilloscope to produce a direct visualization of the audio waveform. I chose this visualization to keep the focus on the audio and music, while giving a nod to early computer graphics that used vector displays. FM audio synthesis with various stereo modulations happens to generate particularly interesting patterns when visualized in this way.