How we make
a twin.

Most "modeled" plugins capture the output of a piece of gear in a snapshot — one knob position, one signal type, one moment in the room. They sound right until you turn a knob.

A TwinGear twin captures the gear itself. Every nonlinearity. Every component drift. Every state transition between a hot tube and a cold one. The model is not a recording of what the gear did once; the model is the gear, rebuilt in software.

A static impulse response is a photograph. A TwinGear twin is a working replica.

A preset is one point in the space of everything your gear can do. Every other setting between presets is interpolation, or worse, a guess.

TwinGear models the parameter space directly. Every knob, every switch, every interaction between them — captured continuously, not sampled. Move a control on the twin and you get the same shift you would hear on the real preamp, including the way that shift interacts with the rest of the signal chain.

Two knobs that fight each other on the original fight each other on the twin. Two knobs that work together, work together. The behavior is not modeled per setting; it is modeled across the space.

Sample rate is not a sticker. Aliasing artifacts hide in the same frequency range as the harmonics that make analog gear feel warm. Train at 48 kHz and the twin learns the artifacts along with the music.

TwinGear runs at 96 kHz from capture to render. The signal we record from your gear is 96 kHz. The model trains on 96 kHz data. The plugin renders natively at 96 kHz internally — and resamples cleanly down to whatever rate your session is set to.

No 48-kHz training data. No resampling shortcut on the way in. The fidelity comes from what we built, not from what your DAW happens to be doing today.