Sam

The 1982 Software Automatic Mouth speech synthesiser. Type text, then play it from MIDI, one word per note sung at the key's pitch, or let the host transport read it aloud, one word per step.

Bundle: sam.vst3
Role: Speech synthesiser

Opinion

Not yet auditioned in a host; the speech itself has been checked only through the tests and by measurement. SAM is the voice of an 8-bit home computer: flat, buzzy and unmistakable, and that is the point. Expect it to be best as a robot lead or a rhythmic spoken part, with Tone pulled down a little to take the edge off its 22 kHz aliasing.

It is a speech synthesiser first: the words are always SAM’s own. The musical part is how they are timed and pitched, and in Keys mode a melody sung on a vowel-rich text sounds more like a vocoder than a singer.

SAM’s source has no licence (it is reverse-engineered abandonware), so this plugin is built from a vendored, patched copy whose redistribution terms are unresolved.

Functionality

Type the text into the box. Words are separated by spaces; a lone _ is a rest. Read it as chooses English, which SAM spells out by its own rules (. falls and ? rises in pitch), or Phonemes, where you write SAM’s codes directly, such as /HEH3LOW2 WER4LD.

Keys mode is a MIDI instrument. Each note speaks one word at the pitch of the key, tuned to within about 12 cents. Next word takes the words in turn and starts again when the host starts playing; Key = word speaks the word a key selects, with C3 the first. Velocity sets level, pitch bend moves the voice two semitones, Release sets the fade when a key comes up, and eight notes can sound at once.

Transport mode is text-to-speech locked to the song. Word n starts on step n counted from Start at (a bar), with steps from a sixteenth note to four bars, following tempo and time signature. Loop repeats the text. Jump the playhead, or let the host loop, and the word under it is spoken from the matching point. It is silent while the host is stopped: Speak plays one pass without the transport, to hear what you have typed.

Both modes use SAM’s four voice controls, Speed, Pitch, Mouth and Throat, and the six voices from its manual (SAM, Elf, Robot, Stuffy, Old Lady, Alien) are one-click presets of them. Monotone sings each word at one pitch instead of letting SAM shape it from stress and punctuation. The current word is highlighted in the box while it plays.

The text is saved with the project. A word is rendered on the audio thread when it starts (a short one takes about 40 microseconds), so nothing is allocated and the timing is sample-accurate.