Mutable Instruments Beads is now open source. Both the hardware design and the code went up on Émilie Gillet’s GitHub (pichenettes) under the MIT license in the early hours of 12 October 2026.
Beads was her last module, and the only one she never opened.
Beads used as a wavetable oscillator — one of several things the module does besides granular processing.
Why this one was different
Mutable Instruments left the Eurorack market in December 2022, nearly four years ago. Gillet’s practice of publishing her designs is a large part of why her modules mattered beyond their own sales: Braids, Clouds, Plaits, Rings, Marbles and the rest are all open, and the result is a decade of community firmware, clones and derivatives that outlived the company.
Beads was the exception. The likely reason, as Synth Anatomy reads it, is timing — releasing the design of a current module while winding the company down would have invited clones of the thing still on shelves. Alternative firmware existed, but the official design did not.
Four years later, the reason has expired.
What Beads actually is
The spiritual successor to Clouds, Mutable’s hugely popular granular processor — but not a Mark 2. A ground-up reinvention, new inside and out.
The concept carried over: a live granular processor for incoming stereo audio that you manipulate on the fly. Gillet’s stated aims for the redesign were a crisper, broader sound palette, more control, better playability, and direct access to new features.
It works in several modes:
- as a delay and reverb
- as a granular processor on live input
- as a granular synth, granulating eight internal banks of wavetables with no input at all
That last mode is the one people forget Beads has, and it is why the video above shows it used as an oscillator rather than an effect.
What happens next is predictable in two directions
Clones. As Synth Anatomy puts it, module cloners will “devour the information and build clones as quickly as possible to get to market first.” Clouds was cloned extensively and Beads will be too. This is the known cost of opening a design and Gillet has accepted it before.
Derivatives. The more interesting direction. Clouds’ open firmware produced Parasite and a family of alternative builds that turned one module into several instruments. Beads has a more capable platform underneath it, and a community that already knows the codebase conventions.
Synth Anatomy’s wish is a good one and worth amplifying: a “Plaits Palette”-style site where the community can collect and share alternative Beads firmware. Plaits has exactly that, and it is the single thing that most extends a module’s useful life — a place where the forks are findable rather than scattered across forum threads.
The observation about who got there first
Synth Anatomy’s editor adds a sharp aside, and it is the part of this story that is specific to 2026 rather than to 2022:
“The first to benefit weren’t humans, but the AI robots. They took it, learned it, and they certainly won’t mention or respect the MIT license.”
And the speculation about motive that follows: “She might have thought: ‘Before an AI tears my code apart, I’ll make it open-source myself.’”
That is speculation and labelled as such. But the underlying point stands on its own, and it connects directly to two other threads this week. The MIT license requires attribution — it is permissive about use and explicit about credit. A model trained on the repository satisfies neither the letter nor the spirit of that, and there is no mechanism by which it would.
It is the same tension Synth Anatomy raised about paid AI-generated firmware mods built on models trained on open-source code, and it will recur every time a significant audio codebase opens. Beads is a high-quality, well-documented, MIT-licensed DSP implementation by one of the field’s best engineers. It is exactly the kind of thing that is most valuable to a training set and least protected by its own licence.
None of which is a reason not to open it. It is a reason the MIT licence now means something different in practice than it did in 2015.