Engram review (2026): an honest verdict on Thoughtful Things' offline AI sampler that turns model glitches into music — sound, workflow, price, and limits.
Engram is one of the most genuinely original instruments to launch in a while: a pad-based sampler and groovebox that runs small generative audio models entirely on the hardware, offline, and lets you deliberately "bend" them into glitchy, uncanny sounds you'd never find in a sample pack. As a sound-design idea it's excellent, and the offline, no-subscription stance is a real differentiator. The honest caveats are big, though — it's a Kickstarter research prototype with provisional specs, a boutique price ($675 and up), and a deliberately niche, experimental audience. It makes audio, not content, so it publishes nothing. Score it as the inventive niche instrument it is, not a mainstream beat maker.
Engram, from Los Angeles startup Thoughtful Things, is a pad-based AI sampler and groovebox that launched on Kickstarter on September 27, 2026. Its founder, Evan King, describes it as "a field recorder for latent space" — an instrument built to go digging for sounds that don't exist until a generative model malfunctions into them. Rather than chase clean, radio-ready AI music, Engram embraces the opposite: it wants the weird, the broken, and the uncanny.
What makes it stand out is that the generative audio models run directly on the hardware, fully offline, with no internet connection and no subscription. Thoughtful Things frames it as the first musical hardware to embed generative audio models in the box itself. Its headline feature, "model bending," is described as circuit bending for neural networks — each model parameter maps to a knob, and pushing them past their intended range deliberately glitches the model's audio decoder into new textures you can sample, slice, and sequence.
This review scores Engram for what it is: a niche, experimental instrument for musicians and sound designers, judged as it stood at launch on 2026-09-27. A disclosure up front — I run Kompozy, a content generation and publishing engine — so I'll be careful not to score Engram against a tool in a different category. Engram makes sound; Kompozy makes and publishes content. Where the two touch is narrow and I'll name it honestly. And because Engram is still described as a research prototype whose design may change, treat every spec below as provisional and reconfirm against the campaign before backing it.
Engram is a hardware sampler and groovebox with generative audio built in. Alongside conventional sampler features — recording and warping your own audio via an on-board microphone or stereo line input, a step sequencer, sample slicing, and on-board FX — it can generate and glitch short bursts of new audio using small models that run locally on the device. Connectivity includes a headphone output, MIDI in/out, and CV sync and gate for integrating with a modular or hardware setup. Thoughtful Things has also described pluggable, shareable audio models, closer to how vintage ROMplers swapped sound ROMs than to a fixed feature set. Its defining choice is philosophical: Engram treats a generative model's failures as a feature. "Model bending" is meant to be explorable, tweakable, and — most importantly — breakable, so the sounds you capture are the degraded, uncanny outputs the model produces when pushed past its limits. Combined with the fully-offline, no-subscription design, that makes Engram less a "make me a song" machine and more a specialized instrument for original, one-of-a-kind sound design.
The clear fit is experimental musicians, beatmakers, and sound designers who want original textures nobody else has and who value owning their tools outright — no cloud, no login, no monthly fee. If you already reach for a Teenage Engineering box, a Chompi, or a Eurorack rig and you're drawn to glitch, granular, and lo-fi aesthetics, Engram is aimed squarely at you, and the CV/MIDI I/O means it slots into an existing hardware setup. Where it fits poorly is anyone wanting mainstream, structured song creation, or anyone who is really after finished content rather than raw sound. It won't write you a polished track the way a text-to-song service does, and it produces no video, captions, graphics, or posts — it is an instrument, not a content pipeline. It's also a crowdfunded prototype at a boutique price, so it's a poor fit for anyone who needs a proven, shipping product today or who wants a cheap entry point into sampling.
| Dimension | Score | Why |
|---|---|---|
| On-device / offline generative audio | 4.6 / 5 | Running generative audio models locally with no internet or subscription is genuinely novel and a real ownership advantage. |
| Sound design & creative range | 4.3 / 5 | "Model bending" opens a distinctive palette of glitchy, uncanny textures that are hard to get any other way. |
| Sampler & sequencer workflow | 3.8 / 5 | Mic and line-in sampling, slicing, a step sequencer, and FX cover the groovebox basics, though depth is unproven pre-release. |
| Connectivity (MIDI, CV, I/O) | 4.0 / 5 | MIDI in/out plus CV sync and gate make it a real citizen of a hardware or modular rig. |
| Value / pricing | 3.4 / 5 | At $675 and up it sits in boutique territory; fair for the novelty, but not a budget entry into sampling. |
| Accessibility & audience breadth | 3.0 / 5 | Deliberately niche and experimental — aimed at sound designers, not mainstream music makers. |
| Maturity & availability | 2.6 / 5 | A Kickstarter research prototype with provisional specs and crowdfunding delivery risk; nothing has shipped at review time. |
| Usefulness for content production | 1.6 / 5 | It makes audio, not content — no video, captions, graphics, or publishing, so it's one input among many for a creator. |
Engram's Kickstarter price starts at $675, described as roughly a 30 percent early-backer discount. Thoughtful Things has not set a final retail price; The Verge estimated it would likely land around $850–$900. That puts Engram in boutique-instrument territory — above budget grooveboxes that run a few hundred dollars (Teenage Engineering's EP-133, the Chompi) and well below a flagship like the OP-1 Field, which sells for over $1,000. For a small-batch hardware instrument with a genuinely novel capability, that's a defensible position, not a rip-off.
The real cost consideration isn't the sticker — it's the crowdfunding risk. Backing a research prototype means paying now for hardware whose design may still change and whose delivery timeline is a projection, not a guarantee. That's normal for Kickstarter and worth pricing in mentally: this is an early-adopter purchase, not a shipping product you can return.
Against that, the offline, no-subscription model is a quiet long-term value point. Many AI audio tools are recurring monthly costs; Engram is a one-time purchase that keeps working with no account and no cloud dependency. For a musician who plans to use it for years, "buy it once and own it" compares favorably to a subscription that never stops — provided the campaign delivers.
| Use case | Fit | Why |
|---|---|---|
| Original, glitchy sound design and sample creation | Strong | This is Engram's core purpose — model bending is built to produce uncanny, one-of-a-kind textures. |
| Owning your tools outright, offline and subscription-free | Strong | On-device models with no internet or account are exactly what Engram is designed around. |
| Integrating with a hardware or modular rig | Strong | MIDI in/out plus CV sync and gate make it fit naturally into an existing setup. |
| Making royalty-free audio to use under your own videos | OK | The samples are yours to use, but Engram makes only the audio — you still need a tool to build and publish the video. |
| Mainstream, structured song creation | Weak | It embraces broken, experimental sound; it won't hand you a clean, radio-ready track like a text-to-song service. |
| A cheap first sampler for a beginner | Weak | Boutique pricing and a "breakable" learning curve make it a poor budget entry point into sampling. |
| Producing finished content — video, posts, graphics | Weak | Entirely out of scope; Engram generates no video, captions, or graphics and publishes nowhere. |
To be clear about lanes: Engram and Kompozy are not competitors, and scoring one against the other on its core job would be a category error. Engram is a hardware instrument for making original sound; Kompozy is a content generation and publishing engine for making and shipping video, images, blogs, and social posts. For sound design, use Engram. For turning ideas into a published content calendar, use Kompozy. Neither replaces the other.
There is exactly one place they touch, and it's worth naming honestly. The samples you bend and record on Engram are original audio you own outright — no license, no content-ID risk — which makes them an ideal, distinctive audio bed for short-form video. If you export a loop from Engram and hand it to Kompozy, it becomes the soundtrack under Persona Shorts, Clipped Shorts, or Marketing Shorts that Kompozy generates, captions per platform, and publishes on a schedule. So the accurate framing is sequential, not competitive: Engram is upstream sound design; Kompozy is where that sound turns into content people actually see.
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. Its signature feature, "model bending," deliberately glitches the on-board model to produce uncanny, original sounds you can sample, slice, and sequence. It launched on Kickstarter on September 27, 2026.
If you're an experimental musician or sound designer who wants genuinely original textures and values owning your tools offline with no subscription, Engram is one of the most interesting instruments to launch recently and likely worth watching or backing. If you want mainstream, structured songs, a cheap first sampler, or a proven shipping product, it's a weaker fit — it's a boutique-priced Kickstarter research prototype aimed at a niche audience.
Kickstarter early-backer units start at $675, described as roughly a 30 percent discount. Thoughtful Things hasn't set a final retail price; The Verge estimated it would likely be around $850–$900. Because it's a research prototype, pricing and specs may change before delivery.
No. Its generative audio models run entirely on the hardware, offline, with no internet connection and no subscription required. Thoughtful Things frames it as the first musical hardware to embed generative audio models directly in the box rather than calling out to a cloud service.
No. Engram is an instrument that produces audio — it makes no video, captions, graphics, or posts and publishes nowhere. To turn its samples into finished, published content, you'd pair it with a content engine: export a loop from Engram and use it as the audio bed under video that a tool like Kompozy generates, captions, and publishes across platforms.
They solve different problems. Suno and similar services generate complete, structured songs from a text prompt in the cloud. Engram is a hardware sampler focused on experimental, breakable sound design that runs offline on the device — its goal is original, uncanny textures you play and sample, not a finished radio-ready track.
The samples you generate and record on Engram are your own audio, so they don't carry the licensing or content-ID entanglements that come with using commercial tracks. That makes them well suited as original, claim-free soundtracks under your own videos. As always, confirm the campaign's own terms before commercial use.