Engram turns AI audio glitches into an experimental instrument
Thoughtful Things launched a Kickstarter for Engram, a sampler and groovebox that uses locally run AI to transform incoming audio and generate unusual sounds. The device is not meant to produce polished songs, but to help users explore experimental and uncanny audio. The company says its models were trained on commercially licensed open datasets and plans to let users modify firmware or load custom models.
Thoughtful Things’ Kickstarter introduces Engram as a sampler and groovebox that runs compact AI models on-device rather than relying on cloud services. It processes incoming audio to create glitchy, uncanny results, and the company positions it as a tool for experimental sound exploration rather than finished commercial tracks.
The firm says its audio models were trained only on open datasets with commercially usable licenses, such as CC-BY or similar. Engram’s firmware is planned to be open for modification or custom model loading. A campaign video shows founder Evan King asking for “piano” and receiving a glitchy result that only loosely resembles one. Early pledges start at $675.
Engram may appeal to musicians, sound designers, and hobbyists interested in experimental audio, offering a local, offline way to explore AI-generated glitches without cloud dependence. Its open-firmware plans could encourage tinkering and custom models, potentially influencing how small creative AI devices are built. At $675+, however, access may remain limited to enthusiasts, so its broader societal effect could be modest, mostly shaping niche music and maker communities rather than mainstream music production.