DECtalk: Klatt's synthesis as a device
In 1983 Digital Equipment Corporation announced DECtalk, a device that turns English text into speech by the rules of Dennis Klatt's laboratory system Klattalk at MIT, licensed in 1982. A digital formant synthesizer builds the speech, and any computer or terminal can drive it.
Why it matters
Rules that had lived in laboratories for decades became a product with several voices, intelligible speech and a telephone connection. Among its uses DEC names computer access by telephone and the visually impaired.
By Klatt's account a Motorola MC68000 processes text a clause at a time and produces synthesizer parameters every 6.4 ms, and a TMS-32010 signal processor produces 10,000 12-bit samples a second; a 6,000-word exceptions dictionary takes about half the memory; output reaches 5 kHz for intelligibility over a loudspeaker. DEC's engineers wrote in April 1983 of a lexicon to grow to about 7,000 words and take over half the read-only memory. The DTC01 owner's manual lists seven voices and one user-definable: Perfect Paul, Beautiful Betty, Huge Harry, Frail Frank, Kit the Kid, Rough Rita, Uppity Ursula and Variable Val; Whispering Wendy appears only in Klatt's recorded examples of 1987. In the modified rhyme test (Logan et al., 1986) DECtalk erred on about 3% of sounds, Prose-2000 about twice as often. The first version's price, Klatt says, came to about four times the early estimate of $1,000, which put it beyond many disabled users; BYTE in February 1984 called it a $4,000 system. The record does not claim the month of the announcement, or whose synthesizer Stephen Hawking used: Digital's journal of 1995 says DECtalk, and no other source was found to check it.