Engram is a sampler that turns broken AI hallucinations into music
Thoughtful Things has launched a Kickstarter for Engram, a hardware sampler and groovebox that runs a locally stored "tiny AI" to distort incoming audio and generate novel, often uncanny sounds. The instrument is geared toward experimental sound design rather than producing polished, commercial tracks and includes open firmware for user modification.
Why It Matters
Engram applies on-device AI to creative audio processing while avoiding cloud-based models and copyrighted training data, signaling a different approach to music tools that prioritizes experimentation and user control. Its open firmware and custom-trained local model may influence how hardware makers balance AI capabilities, data ethics, and user remixability.
Key Facts
- Product: Engram sampler and groovebox
- Company: Thoughtful Things
- Launch platform: Kickstarter campaign
- AI model: Locally running "tiny AI", designed in-house and custom-trained
- Training data policy: Trained on open datasets licensed for commercial use (CC-BY or similar); company says it will not train on non-commercial, pirated, or stolen data.
Music technology startup Thoughtful Things has opened a Kickstarter campaign for Engram, its first hardware instrument. Engram combines sampling and groovebox features with an embedded, locally running AI the company describes as a "tiny AI." Rather than connecting to cloud services, the device processes audio on-device to mangle input and to hallucinate new, often glitchy sounds. The instrument is positioned for experimental sound design rather than mainstream song production. In the campaign video founder Evan King demonstrates prompting Engram by voicing the word "piano," then receiving a warped, piano-like output. Thoughtful Things frames the device as a "field recorder for latent space," inviting users to explore AI-generated artifacts and unexpected timbres rather than aiming for polished, radio-ready results. Thoughtful Things says it trained Engram's audio models on open datasets licensed for commercial use (for example, CC-BY or similar) and asserts it will not use non-commercial or pirated material for training. The company also plans to publish Engram's firmware, allowing users to modify the software or load custom models; the design takes cues from circuit bending by encouraging hands-on tweaking and even "breaking" the tiny AI models to discover new sounds. The Kickstarter lists a limited initial run of Engram units, with pledge pricing starting at $675. The company has not confirmed a final retail price on the campaign page, though the $675 pledge is presented as a 30 percent discount and suggests a likely retail range between $850 and $900.
Keep Reading
Out of the Park Baseball lets me enjoy baseball even when the Mets suck
Can Muse overcome Meta’s trust issues?
Waiting at a red light? It may not know you’re there
