In this mode, the incoming signal is converted into “frames” of spectral data, that are stored, transformed, recombined, and resynthesized as a time-domain signal.
POSITION selects into which buffer the audio is poured (when FREEZE is not active), or from which buffer the audio is synthesised (when FREEZE is active). For example, set POSITION to its minimum value. Press FREEZE. You get a first texture. Set POSITION to its maximum value. UNFREEZE. Wait for something else to happen in the incoming audio. Press FREEZE again. By moving POSITION you interpolate between the two textures which had been captured at the press of FREEZE. Depending on the quality settings there are 2 to 7 buffers laid out over the course of the POSITION knob. What the module does is crossfade between a “wavetable” of FFT slices.
SIZE changes the coefficients of a polynomial that determines how frequencies are mapped between the analysis and synthesis buffers. It’s like a 1-knob GRM Warp. Over the course of the knob it’ll do spectral shifting, but also spectral reversal.
PITCH controls the transposition (pitch-shifting).
DENSITY determines how results from the analyzer are passed to the resynthesizer. Below 12 o’clock, there’s some increasing probability that a given FFT bin won’t get updated, causing a kind of partial freeze. After 12 o’clock, adjacent analysis frames are increasingly merged together (like a low-pass filter in the amplitude each frequency bin). At extreme settings, random phase modulation is applied to smooth things - giving you different flavours of spectral muddling/reverb.
TEXTURE does two things: below 12 o’clock, it increasingly quantizes the amplitudes of the spectral components, like a very low-bitrate audio file. After 12 o’clock, it increasingly weakens the strongest partials and amplifies the weakest ones. This has the effect of making the spectrum more noise-like.
IN: Sets the input level
OUT: Sets the output volume
PAN: Sets the amount of random panning/balance applied to the grains. Only noticeable with stereo output.
VERB: Adds reverb to both wet and dry signal.
FB: Amount of signal feedback. Watch out because things can get out of hand with feedback in this mode. If you have tons of crazy feedback happening and the pedal seems like its getting away from you, just turn FB all the way down. I usually leave it all the way down.
Sending a trigger to the TRIG input creates various frequency-domain glitches typical of corrupted (encoded) audio files. It works as a kind of build-up/feedback effect - the shorter the pulse, the smaller the effect. With a continuous gate, it’ll really start to rot…