SAM is a formant speech synthesizer in a Windows 64-bit VST3 shell. Internally it's the same engine that ships built-in with Harmonic DAW — the in-DAW instrument extracted, repackaged, and exported as a standalone VST3 so you can drop it into Cubase, FL Studio, Reaper, Ableton, or anything else that loads VST3. Its lineage runs back to Don't Ask Software's 1982 Software Automatic Mouth — the same SAM that gave the Commodore 64 a synthetic voice — and through the formant-rebuild SAM2600 ports of the '90s. Same character, same boxy lo-fi vowels, same robotic intent. Rebuilt from scratch in JUCE 8 with eight-voice polyphony, MIDI-driven glottal pitch, and a singing mode that lets vowels sustain for as long as you hold a note.
The signal path is bone-simple by design. SAM runs internally at 22050 Hz — the classic SAM sample rate — and linear-interpolation upsamples to whatever your project rate is at the output stage. That fixed internal rate is responsible for the boxy lo-fi character you hear even on a 96 kHz session: every formant filter, every glottal pulse, every noise burst is computed at the original SAM resolution. Three resonant band-pass filters in parallel shape the spectrum into vowel-and-consonant formants — F1, F2, F3 — with the phoneme rule table moving them per letter. The excitation is either a glottal pulse train with a touch of breath noise (for voiced sounds — vowels, m, n, r, l, v, z) or a high-pass-filtered white noise burst (for unvoiced sounds — s, f, t, k, p). Phoneme cells transition hard, with no formant interpolation. That hard cutover is what gives SAM its choppy, almost glitchy diction — the audible seam between a vowel and a consonant that everyone associates with the sound of an early-'80s talking computer.
Type words into the text field, send MIDI notes, and each note-on triggers the spoken text at that pitch. The glottal pulse frequency tracks the note (A4 = 440 Hz, exactly like a synth oscillator), so speech can be played as a melodic instrument — chords, arpeggios, harmonic stacks, all carrying the same robotic articulation across every voice. Up to eight notes at once, each holding its own snapshot of the text taken at the instant of note-on, so you can type while the keyboard is being played without mangling in-flight utterances. The PITCH knob is an override: leave it at zero and the voice follows the MIDI note; nudge it off zero and it pins the glottal frequency for monotone narrator effects. SINGING loops the final phoneme cell so vowels hold for the full duration of a sustained note instead of cutting off when the text runs out — useful for melismatic lines, drone vocoder textures, and any pad where you want the last word to last.
Pay with PSYKO — the native token of the Psykovsky ecosystem on the BNB Smart Chain. Connect any BEP-20 wallet (MetaMask, Trust, etc.), confirm the transaction, and the plugin download begins automatically. No accounts, no email confirmations, no refund department.
PRICE
$5
— PAID IN PSYKO ON BNB SMART CHAIN —