Thoughtful Things launches Engram AI music sampler

Thoughtful Things has introduced Engram, a hardware sampler and groovebox that leverages offline artificial intelligence to manipulate audio and generate uncanny, experimental sounds. Unlike mainstream generative music tools that aim to produce polished, radio-ready tracks, Engram is built for sound designers who want to push machine learning models to their absolute limits. The device operates entirely without an internet connection, running a lightweight, custom-trained AI model directly on its internal hardware.

To address ethical concerns surrounding generative media, Thoughtful Things trained its proprietary audio models exclusively on open datasets containing commercially licensed audio, such as Creative Commons material. The startup has committed to keeping its training pipeline free of pirated or unlicensed data. Furthermore, the company plans to open-source Engram's firmware, allowing musicians and developers to modify the system or upload their own custom-trained models to the hardware.

In practice, the device acts as what founder Evan King calls a "field recorder for latent space," enabling users to input audio or voice prompts and receive glitchy, abstract interpretations. For example, prompting the device for a piano sound might yield a heavily distorted, vaguely recognizable acoustic texture. This approach allows electronic musicians to circuit-bend the underlying AI, intentionally breaking the model to discover unique sonic artifacts.

The hardware is currently available through a Kickstarter crowdfunding campaign. Thoughtful Things is offering an initial limited run of Engram units starting at $675, which represents a 30 percent discount off the expected retail price. Once the campaign concludes and the device enters standard production, the retail price is projected to land between $850 and $900.