Hermetic Modular’s Capicola is a 12HP Eurorack module that uses a keyframe-based approach to stretch, pitch-shift, slice and distort stereo audio in real time.
Two things in that sentence are harder than they look, and one is genuinely novel for the format.
Why real-time time-stretching is hard
Changing a sound’s duration without changing its pitch — or its pitch without its duration — is not a simple operation, because the two are physically coupled. Play a tape faster and both change together. Decoupling them requires reconstructing the signal.
The three standard approaches, each with a characteristic failure:
Granular / overlap-add (OLA, WSOLA). Chop the audio into short grains and re-space them — overlap them to compress, spread them to stretch, crossfading between. Cheap, robust, and it produces the familiar granular smearing: a slight warble on tonal material and a stuttering on transients, because grain boundaries do not respect the waveform’s period. WSOLA improves this by searching for the best alignment within a window.
Phase vocoder. Take an STFT, scale the time axis, and resynthesise — adjusting the phase of each bin so partials stay continuous. Much better on sustained tonal material, and it produces the characteristic phasiness: a loss of transient definition and a slightly smeared, underwater quality, because the phase relationships between bins that define a sharp attack are exactly what the algorithm has to tamper with.
Sinusoidal modelling / spectral. Track individual partials and resynthesise. Best quality on harmonic material, expensive, and it fails on noise.
And real-time makes all three harder because you cannot look ahead. An offline stretcher knows what is coming and can plan grain boundaries or phase corrections around transients. A live one is working with whatever has arrived.
In 12HP — roughly 61mm of panel — with stereo throughput, on an embedded processor. That is a tight budget for any of the above.
Why keyframes are the interesting choice
Here is where it departs from Eurorack convention.
The normal modular answer is a CV input. You want time-varying stretch, you patch an LFO or an envelope into a stretch jack, and the parameter follows the voltage. That is the format’s whole idiom: everything is a voltage, modulate it with another voltage.
Keyframes are a different model. You specify values at points in time and the system interpolates between them. It is the animation and DAW-automation paradigm, not the modular one.
What that buys you:
Repeatability. A CV-driven stretch is whatever the LFO happened to be doing. A keyframed one is the same every pass — which matters enormously if you are building a part rather than exploring a texture.
Non-monotonic, shaped trajectories. Getting a stretch that ramps slowly, holds, snaps back and then accelerates requires a complicated modulation patch with CV. With keyframes it is four points.
And coordination across parameters. Keyframing stretch and pitch and slice position together, in a known relationship, is a composed gesture. Doing that with three correlated CV sources is a patch in itself.
The cost is the obvious one: keyframes are less jammable. The joy of modular is patching something and discovering what it does. A keyframe editor on a 12HP panel is an interface you program rather than play, and the quality of that interface — how you enter, see and edit the points on presumably a small screen and a couple of encoders — will decide whether the module is used or admired.
This is the same tension we wrote about with the Adamsynths Warthog’s total recall and the Buchla 200e’s preset memory: bringing memory and determinism into a format built on the unrepeatable. It keeps happening because performers keep needing it.
Where it fits
Time-stretching in a modular case is most useful for the things that are awkward anywhere else:
- Sampled material that has to lock to a patch’s tempo without a DAW
- Extreme stretch as a texture generator — the Paulstretch territory, where a two-second sample becomes two minutes of drone and the artefacts are the sound
- Slicing with independent pitch, which is the classic breakbeat manipulation and genuinely hard to do in Eurorack
- And stereo throughput, which is rarer in Eurorack than it should be and matters if you are processing a full mix rather than a single voice
No price or availability is given in the coverage we have seen, so treat this as an announcement rather than a review.