An offline AI sampler and groovebox from Thoughtful Things that runs small generative audio models on the hardware and lets you "model-bend" them into glitchy, original sounds — no internet, no subscription.
Last verified · 2026-09-27 · by Moe Ameen
Engram is a pad-based AI sampler and groovebox from Thoughtful Things, a Los Angeles startup founded by Evan King. It launched on Kickstarter on September 27, 2026, with early-backer units starting at $675. What sets it apart from an ordinary sampler is that it embeds small generative audio models directly in the hardware and runs them entirely offline — no internet connection and no subscription — which the company frames as a first for musical hardware.
Its signature feature is "model bending," described as circuit bending but for neural networks. Each parameter of the on-board audio model maps to a knob, and turning them past their intended range deliberately breaks the model's audio decoder, producing glitchy, degraded, uncanny textures. King calls the device "a field recorder for latent space" — a tool for going digging for sounds that don't exist until the model malfunctions into them. Rather than chase clean, radio-ready output, Engram treats a model's failures as the creative material.
Around that core it's a capable groovebox: it samples your own audio through an on-board microphone or stereo line input, generates and glitches short bursts of new audio, and lets you slice, sequence, and add FX. Connectivity includes a headphone output, MIDI in/out, and CV sync and gate. Thoughtful Things has also described pluggable, shareable audio models, ROMpler-style. Note that Engram is still described as a research prototype whose design may change, and no final retail price is set — The Verge estimated roughly $850–$900 — so treat specs as provisional.
Think of Engram as the front of a two-stage pipeline whose second stage is distribution. On its own, Engram gives you something most creators struggle to get: a sound that is unmistakably yours and belongs to no one else. The problem is that a folder of brilliant, bent one-shots does nothing for an audience until it's the audio under video that ships — and ships often. That's the gap [Kompozy](/) closes, and the most useful way to use the two together is at batch scale, not one clip at a time.
Here's the concrete workflow. Bounce a small "sound pack" off Engram — a hook loop, a couple of transitions, a stinger — then use those files as the audio bed inside Kompozy while it fans a single source into a whole batch of video: [Persona Shorts](/glossary/persona-shorts) talking-head clips, [Clipped Shorts](/glossary/clipped-short) cut from long-form footage, Listicle Video and Naturalistic Video cards over portrait clips, and Marketing Shorts. Because Kompozy is a generation and publishing engine — not a single-clip editor — the same Engram sound threads through the entire run, giving the series a consistent sonic identity the way a brand-exact [HyperFrames](/glossary/hyperframes) template gives it a consistent visual one. Kompozy then writes per-platform captions and, via [Autopilot](/glossary/autopilot), schedules and publishes the batch across the eight social platforms plus your blog and a Mailchimp newsletter. So one afternoon of sound design on Engram doesn't score one video — it scores a week of them, published everywhere, with audio nobody can content-ID against you. If you'd rather generate the music in software instead, Kompozy pairs the same way with [Sonilo](/ai-tools/sonilo), [Suno](/ai-tools/suno), or [Treblo](/ai-tools/treblo); Engram is the option that gives you a sound that's genuinely one of a kind.
Engram is a pad-based AI sampler and groovebox from Thoughtful Things that embeds small generative audio models directly in the hardware and runs them fully offline — no internet or subscription. Its "model bending" feature deliberately glitches the on-board model to create uncanny, original sounds you can sample, slice, and sequence. It launched on Kickstarter on September 27, 2026, from $675.
No. The generative audio models run on the device itself, offline, with no account or monthly fee. Thoughtful Things frames Engram as the first musical hardware to embed generative audio models directly in the box rather than calling a cloud service — so it keeps working with no connection.
Yes — the samples you create on Engram are original audio you own, which makes them ideal, royalty-free soundtracks for short-form video with no content-ID risk. Engram makes only the audio, though. To turn it into finished, published video you'd pair it with a content engine like Kompozy, which generates the video, adds captions, and publishes it across platforms.
Engram is upstream sound design; Kompozy is downstream content and distribution. Bounce a loop or one-shot off Engram, then use it as the audio bed inside Kompozy while it fans a source into Persona Shorts, Clipped Shorts, Listicle Video, and Marketing Shorts — captioned per platform and published on a schedule via Autopilot. One sound pack can score a whole batch of clips at once.
Kickstarter early-backer units start at $675, described as roughly a 30 percent discount. There is no final retail price yet; The Verge estimated around $850–$900. Engram is still a research prototype, so pricing and specs may change before it ships.