Introduction

Engram is a standalone sampler and groovebox that uses a tiny, locally running AI model to reinterpret audio input as experimental, often glitchy soundscapes. Created by Thoughtful Things, the instrument is designed for musicians who want to explore the uncanny side of machine-generated audio without internet connectivity.

What Happened

Thoughtful Things launched a Kickstarter campaign for Engram, a hardware device that processes incoming audio through a custom-trained AI model that can hallucinate entirely new timbres. The system runs entirely on-device, features circuit-bending-inspired controls that let users intentionally break or tweak the model, and was demoed by founder Evan King who asked for piano and heard a glitchy, vaguely piano-like response.

  • Runs a tiny AI model locally, no internet connection required
  • Uses circuit-bending-style tweaks to manipulate the model's output
  • Demo video shows voice-triggered sound generation

Why This Matters

Engram signals a shift in how creators can engage with AI audio tools, moving away from automated track generation models toward hands-on instruments that push models beyond their intended limits. By training on licensed audio data and planning open firmware, the project aims to avoid copyright pitfalls while giving artists granular control over AI-driven sound design.

Key Takeaways

  • Engram starts at $675 on Kickstarter, with expected retail between $850 and $900
  • The company plans to open-source the firmware for custom models and tweaks
  • Training data is limited to commercially licensed audio, avoiding pirated sources

Conclusion

Engram is not intended as a songwriting shortcut or a way to generate radio-ready tracks. Instead, it offers a tactile, experimental pathway into AI hallucinations, giving creators a new tool for sound design and avant-garde music. Its Kickstarter reception may indicate whether there is a lasting market for standalone AI instruments that prioritize artistic exploration over convenience.