Engram Launches AI-Powered Sampler Converting Hallucinations to Experimental Music

Music technology startup Thoughtful Things has launched Engram, a novel sampler and groovebox that leverages AI to reinterpret audio hallucinations as experimental music. Unlike mainstream AI music generators, Engram focuses on creating uncanny, glitchy sounds by intentionally pushing AI models beyond their conventional outputs. The device operates offline, running a custom-trained 'tiny AI' model locally, avoiding cloud dependencies entirely.

Ethical AI and Open Firmware

The company emphasizes ethical data practices, training its models exclusively on commercially licensed open datasets (CC-BY or similar). Founder Evan King describes Engram as a 'field recorder for latent space,' inviting users to explore AI's creative potential through circuit-bending-inspired experimentation. Notably, the firmware will be open-sourced, enabling developers to modify or deploy custom models.

• Locally-run AI model for offline audio manipulation
• Custom-trained on ethical, commercially licensed datasets
• Open firmware for community-driven model customization
• Inspired by circuit bending and experimental sound design

The Kickstarter campaign offers Engram at a 30% discount ($675), with retail pricing expected between $850-$900. Early adopters can access the device while supporting its open-source development roadmap.

Why it matters for GPU/AI infrastructure: Engram demonstrates how edge AI can deliver specialized creative tools without cloud reliance, highlighting the potential for compact, efficient GPU-accelerated models in niche applications like music production and generative art.