Sounds like: Stuttering, vowel-like blips that sound like chopped robot speech.
A rhythmic, stuttering formant-filtered texture that suggests chopped, robotic vocal syllables — built from a fast-gated bandpass sweep rather than any actual vocal source.
Open this patch in the interactive PatchpadALGO 8 with MIX -64 selects the X output (Operator A modulating Operator C, plus B2 as a plain carrier) for a bright carrier. FLTR TYPE at 64 (the LP/BP/HP morph's midpoint) plus high RESO gives a peaky, formant-like resonant band. The 'vowel-jump' movement between gated notes is done by p-locking FLTR page 1's FREQ to a different value per step (a real, genuine Digitone II sequencing technique) -- program 3-4 different FREQ p-locks across the pattern's steps for distinct 'vowel' positions rather than a single static cutoff.
Syntakt's digital-voice tracks only offer Lowpass/Highpass/EQ/OFF, so RESO is pushed high on a Lowpass to approximate a peaky, formant-like band.
LFO 1's destination is hardwired to filter cutoff via its own section's Depth knob, so this route needs no Mod Matrix entry. LFO 1's Square waveform gives a two-position stepped jump rather than a smooth sweep -- the closest real Summit equivalent to a stepped/rhythmic vowel-jump LFO.
Osc 1 is Formant: Wave sets the interval between the two formants, Timbre the base formant frequency (the vowel position) and Shape the formant width. LFO 1 runs a Sqr wave at Sync 1/16. Mod Matrix: the LFO 1 row to the Timbre 1 column at +70. Values are on MiniFreak's native 0-100 scale.
M1's real StepSequencer mode is a genuine per-step modulation source, with its own '#' step count and per-step '->LVL' values, routable directly to CUT. The 4 example steps shown here alternate between low and high cutoff values for a real vowel-jump pattern; program up to the full step count for more variety. Very short AMP ADSR values give the gated-note character.
Filter 2 (the real 2-pole state-variable filter with a continuous LP-BP-HP MORPH) gives a true Band Pass character via MORPH=64 (Filter 1's BP types have no continuous morph). LFO 1 is pre-wired to the filter's own LFO 1 AMOUNT knob, so no Mod Matrix entry is needed; its Square waveform, tempo-synced, gives the stepped vowel-jump movement between gated notes.
This is built entirely from a synth voice, not a real vocal source or vocoder — the 'robotic vocal' quality comes from rhythmic bandpass movement on very short gated notes, achievable on every synth here via sequencing.