IBM 7094 Text-to-Speech: How Computers Learned to Speak

The Voice of a Giant: How the IBM 7094 Learned to Speak and Sing

Long before smart speakers could read the news, summarize emails, or chat with you in natural voices, computers were massive, room-filling mainframes that communicated primarily through blinking lights and clacking punch cards.

That changed in the early 1960s at Bell Labs, where researchers achieved a monumental milestone in computer science: making a machine talk and sing. At the heart of this breakthrough was a powerful vacuum-tube and transistor-based mainframe—the IBM 7094—and a piece of history that would eventually inspire one of the most famous scenes in cinematic science fiction.

The Technology: Simulating Human Speech

In 1961, generating audio entirely through software was far too computationally intensive for the hardware of the era. Computers lacked sound cards, digital-to-analog audio chips, and modern storage.

To overcome this, researchers John Larry Kelly Jr. and Carol Lochbaum, aided by programmer Max Mathews, used an IBM mainframe (the IBM 7090/7094 series) coupled with a hardware vocoder.

• The Physics of Speech: Instead of playing back a pre-recorded human voice, the computer mathematically simulated the human vocal tract.
• Control Parameters: The program directed the vocoder to change pitch, tongue position, and throat resonance over time, effectively synthesizing phonemes (vowels and consonants) from scratch.

“Daisy Bell”: The First Computer Song

While the primary goal was general text-to-speech research, the team wanted a striking demonstration of the system’s capabilities. They decided to program the computer to sing.

Max Mathews arranged the musical accompaniment, while Kelly and Lochbaum coded the vocal track for the 1892 popular song “Daisy Bell” (often known as “Bicycle Built for Two”).

When the IBM 7094 fired up its text-to-speech routines and “sang” the lines:

“Daisy, Daisy, give me your answer do. I’m half crazy, all for the love of you…”

…it marked the first time in history a computer had ever voiced a song. The resulting audio was robotic, eerie, and shaky, but it sounded unmistakably human—a chilling window into the future of technology.

From Bell Labs to 2001: A Space Odyssey

This historic demonstration left a lasting mark on popular culture, largely thanks to science fiction writer Arthur C. Clarke. Clarke happened to be visiting Bell Labs around the time of the experiment and was deeply fascinated by the singing computer.

Years later, Clarke worked with director Stanley Kubrick to write the landmark 1968 film 2001: A Space Odyssey. In the movie’s climax, the rogue supercomputer HAL 9000 is slowly deactivated by the astronaut. As its consciousness fades, HAL regresses to its childhood and sings the exact song the IBM 7094 performed years earlier: “Daisy, Daisy…”