VOICE SYNTHESIS
Wanna go full robot? Lose the singer, get typing and generate robotic vocals from scratch
Use in music
Way back in the 30s, Homer Dudley invented the VODER (Voice Operation DEmonstratoR). In the hands of a skilled operator, this electronic device could produce understandable results – as heard at bit.ly/Voder1939. White-coated boffins continued to experiment, and by 1961, Bell Labs had coaxed an IBM 704 computer into singing Daisy Bell ( bit.ly/DaisyBell).
In the mid 70s, speech synthesis appeared in consumer devices, and the 1976 Speech+ speaking calculator used the first ever speech chip, the TSI S14001A. Get an instant overview at bit.ly/TSIchip.
It was Texas Instruments’ 1978 Speak & Spell that captured the imagination of the public, though, thrusting computer speech into the mainstream. At the time, few had heard an electronic device produce speech on its own, but now you could actually own one! Speak & Spell used ‘Linear Predictive Coding’ (LPC) to reproduce recognisable speech with meagre resources.
As for synthetic speech in music, Kraftwerk were quick on the uptake, with several examples on their 1981 album Computer World, such as Numbers ( bit.ly/KraftNumbers). More recently, good-quality synthesised singing has become viable, and insanely popular in Japan through Yamaha’s Vocaloid series.
How it works
LPC-based synthesis uses an oscillator to model the output of the larynx, noise to represent sibilant sounds, then filters to recreate the formants of a human voice. Another method is concatenation synthesis, where sampled sections of real spoken vocals are strung together to reproduce the desired words. Or how about using additive synthesis to build a voice from scratch using sine waves? There are many voice synthesis methods, in fact, each with its own pros and cons.
Get the sound
To transform your own voice into a computerised one, try Sonic Charge’s Bitspeek or iZotope’s VocalSynth in Compuspeech mode. These break a voice down in real time, then reconstruct it, LPC-style. To create speech and singing from nothing, try Plogue’s chipspeech, which emulates retro voice chips (also check out the free Alter/Ego, which offers a more realistic, modern sound), Yamaha’s Vocaloid, and VirSyn’s Cantor.