The history of computer text-to-speech synthesis

by Evan Koblentz

Does your computer talk? Or rather, does it talk any better than it could approximately 25 years ago?

That's right: we're "talking" about Software Automated Mouth, better known as SAM, developed mostly by Mark Burton in 1979. The company, SoftVoice, still exists today at http://www.text2speech.com. The story of how Steve Jobs used SAM to make the Macintosh computer "introduce itself" in 1984 is detailed at this very long web address: http://www.folklore.org/StoryView.py?project=Macintosh&story=Intro_Demo.txt&topic=Marketing&sortOrder=Sort+by+Date&detail=medium&showcomments=1 (copy and paste the link because it will break across lines) but we prefer http://homepage.mac.com/vectronic/appleii/sam.html where you can actually download the stuffed software for Apple II computers! (Soon we're acquiring a //c and looking forward to getting this great memory from the past.) There is also a Wikipedia entry at http://en.wikipedia.org/wiki/Software_Automatic_Mouth although we don't particularly trust the accuracy of Wikipedia entries.

SAM is very neat, but we wondered: what came before it? What is the real history of computer text-to-speech synthesis? In short, what was the first machine -- computer or not -- to speak for itself? A-Googling we went in search of some answers.

We found the web site http://www.ling.su.se/staff/hartmut/kemplne.htm where Prof. Hartmut Traunmüller reports: "The first attempts to produce human speech by machine were made in the 2nd half of the 18th century. Ch. G. Kratzenstein, professor of physiology in Copenhagen, previously in Halle and Petersburg, succeeded in producing vowels using resonance tubes connected to organ pipes (1773)."

But there is a difference between the first attempts at something and the first actual success. The professor continues: "[Wolfgang] Von Kempelen's machine was the first that allowed to produce not only some speech sounds, but also whole words and short sentences," described in a paper by Von Kempelen in 1791.

Kempelen, as many computer history buffs already know, is more famous for building a supposedly automated chess-playing machine, described in great detail by writer Tom Standage in his 2002 book, "The Turk". That chess machine turned out to be a fraud -- a short human was always hidden inside -- but the speaking machine was real. Standage explains how telegraph pioneer Charles Wheatstone built a copy of the machine in 1863 and demonstrated it to a young Alexander Graham Bell. (Visit http://www.ling.su.se/staff/hartmut/farkas.htm for many more web sites and biographical references about Von Kempelen.)

Jump back to 1979: Texas Instruments was one of a few companies selling handheld language translator devices. TI engineers understood, however, that merely reading a foreign language was useless if you didn't know how to pronounce the word. So they built speech synthesis into the product using off-the-shelf technology from their own parts bin: anyone remember the "Speak & Spell" toy? TI used the toy chip while Burton and his colleagues were working on software solutions! There are some fascinating specifications and other details at http://www.datamath.org/Speech/LanguageTutor.htm. (I used to own one of these devices, but gave it to VCF chief Sellam Ismail. Unfortunately I did not record any audio clips from it.)

So again, we ask: does your computer talk any better than it could approximately 25 years ago?