Orrery — Copenhagen

OR-0032

Salo Oy
Salo, Finland · 1991 – c. 2004

Ser. VX-9640-0088
1996 · effect

Issued as VST3 · AU

Restored 2025 from 1996 hardware.

Salo 6210 — Vocoder

Salo 6210 · analysis/synthesis voice processor · restored from unit no. VX-9640-0088 · Salo Oy, Salo, Finland, 1996

The restorer's note

The robot voice, from the company that already owned the mathematics of speech. Salo built it for secure comms; musicians would have loved it, had they ever been let near one. Serial 0088 — an early, low-numbered unit, restored with its bands untouched.

Acquisition

The Salo Collection $49 · €45 · £39 · 329 DKK The Complete Catalogue $149 · €135 · £115 · 999 DKK

Prices are exclusive of tax. Customers in a country that charges VAT or sales tax will be liable at the rate set by that country; the relevant amount is added at checkout to the purchase price.

Borrow it from the archive for a short time, or acquire it for your collection.

Runs in your studio as AU and VST3 on macOS, and VST3 on Windows.

Download — macOS: installer · zip · Windows: installer · zip

Issue 1.1.0 · 2026-10-08

Earlier issues

Issue 1.0.2 · 2026-08-28 — macOS: installer · zip · Windows: installer · zip

The plate

The restored Salo 6210 — Vocoder: its faceplate as issued.
Plate 32 · as issued

The controls

The controls.

Mode Channel as issued
Chooses the vocoding method: Channel, a bank of band filters, or LPC, the linear-predictive core.
Bands 8–32 · 16 as issued
Sets the number of analysis bands, from 8 to 32 — more bands, clearer speech.
Carrier Internal as issued
Chooses the excitation source the vocoder sings through: Internal, which follows the note played; Sidechain; or Noise.
Formant -12.0–12.0 st · 0.0 st as issued
Shifts the centre frequencies of the vocoder's bands, in semitones.
Attack 0–100 ms · 5 ms as issued
Sets the envelope follower's attack — how quickly it tracks a rising level, and how much it smears.
Release 1–500 ms · 40 ms as issued
Sets the envelope follower's release, after the level falls.
Sibilance 0–100 % · 40 % as issued
Lets unvoiced, high-frequency consonants pass straight through, for intelligibility.
Input -24.00–24.00 dB · 0.00 dB as issued
Sets the level going in, in decibels, before the signal reaches the restored circuit.
Output -24.00–24.00 dB · 0.00 dB as issued
Sets the level coming out, in decibels, after the restored circuit and the Mix blend.
Mix 0–100 % · 100 % as issued
Blends the restored circuit against the untouched signal at equal power, from the untouched signal alone at 0 % to the circuit alone at 100 %.
Bypass Off as issued
Takes the restored circuit out of the path and passes the untouched signal, delayed to match the circuit so nothing shifts in time.

A stored setup is a memory. MENU opens the list: the display reads MEMORY 03/12, with the memory's name below. ◂ ▸ step through it without changing the sound, and OK loads the one shown. Holding OK asks STORE? OK; OK again opens the name, which starts as USER NN. ◂ ▸ change the letter, OK moves on, and holding OK stores it; the computer keyboard may type the name instead, Return storing and Escape cancelling. Holding MENU on a user memory asks CLEAR? OK.

Listening

Listening

Demonstration recordings, prepared at the bench.

  1. The voice, played

    The note reads dry for six seconds; then the Internal carrier takes it — a held line sung by the filterbank, sixteen Bands wide — and the reading becomes the machine's.

    0:00 / 0:30

  2. Eight bands to thirty-two

    Bands step from eight — barely words — to sixteen, then thirty-two, the reading sharpening at each move: the resolution of the filterbank, heard arriving. Mode then switches to LPC, and the same voice is carried by the all-pole model instead.

    0:00 / 0:30

  3. Spoken through the pad

    Carrier switched to Sidechain: the pad is made to speak. Sibilance rises to 80 % and the consonants return; Formant is pushed up seven semitones, then down; Attack and Release open at the close and the words smear.

    0:00 / 0:30