Engram seeks backing for an offline AI sound sampler
Thoughtful Things opened a Kickstarter campaign for Engram, a sampler and groovebox built around a small audio model that runs on the device. It can transform recorded sound or generate brief samples for a musician to edit and arrange. The company's account of how it trained that model has not been independently audited in the retained reporting. The campaign is for an instrument still awaiting completed deliveries.
Artificial Intelligence··Night
Thoughtful Things opens crowdfunding for Engram
A new crowdfunding drive by Thoughtful Things seeks to produce Engram, the company’s first instrument for making music. The hardware holds a sampler, a beat sequencer and a compact audio model; its sound generation happens locally. A musician can feed it a recording, alter that sound, or generate a brief new sample to use in a composition. The model does not require an internet connection. The company presents the instrument as a means of exploring unusual sounds, rather than a system that delivers a finished song from one instruction. The campaign concerns an instrument being offered to backers, not evidence that completed units have reached customers.[1]
How the audio is shaped
Engram is built for manipulating small pieces of audio. In a campaign demonstration described by The Verge, founder Evan King says the word 'piano' into the device and receives a glitchy sound resembling a piano. The musician can then sample and arrange that output. The concept borrows from circuit bending, in which changing how a device operates produces unexpected sound. Here the controls push a small audio model away from ordinary output. Thoughtful Things intends to make the firmware open so users can adjust the device or load their own models. That is a stated plan for the software, not a completed release described by the report.[1]
Company says it trained the audio model itself
Thoughtful Things says it designed and trained the audio model itself. It says its training material came from open audio datasets licensed for commercial use and excluded pirated or noncommercial material. Those are company statements; the retained reporting does not describe an independent audit of the datasets. The campaign is for a limited first production run, and no retail price has been announced. No outside listening assessment or record of finished units reaching buyers appears in the retained evidence. For now, the concrete development is the opening of crowdfunding for a locally running sound instrument, with its training provenance and eventual delivery still resting on the maker's account and plans.[1]
