The Making of Documentary

In 1986, you could put a moose on a Macintosh screen. He talked aloud and told jokes, with his mouth moving along with the words.

The Talking Moose was created by Canadian physician and programmer Steven B. Halls, at a time when a recognizable face had to be built out of black-and-white pixels. It didn't take many pixels to give this one a personality.

Questions about artificial intelligence tend to follow today's talking computers. A different sort of story covers the Moose's beginnings: one programmer figuring out how a computer-generated voice could seem to come from a character.

Why a moose?

Antlers helped.

Photographs of the original Talking Moose present a two-dimensional Macintosh character. With so little room for detail, its outline had to carry most of the work. Antlers, Halls recalls, were easy to make out at a small size. He also happened to like moose, so the choice was settled.

The first version was written by Halls in assembly language. Making the lips follow the speech took real ingenuity. As he recalls it, the program located Macintalk, the Macintosh speech engine, in memory. It watched information about the sound being played, then adjusted the Moose's mouth shape accordingly.

This didn't change the speech engine's memory. The Moose was watching it. Accessing memory across applications was possible in early Macintosh software, though modern operating systems generally restrict it.

The character could also comment on window and document names in familiar applications, thanks to that access. When the Moose seemed to know something about your work, it was responding to information on your own computer. Its behavior was programmed, without a language model behind the face.

The Years in Color

Around 1989, the Moose gained color. A picture of that version, printed with an early Macintosh color printer, hung on Halls's wall. His account of the color Moose includes scans where the printing dots are still visible.

It is a small piece of computing history you can inspect, the character as it looked then, preserved on paper as well as in software.

Another kind of record came from old users. In a letter published in 2016, Mike Ellis remembered seeing the Moose at a Boston Mac Developers Group conference in the late 1980s:

"He was widely admired by some of the cleverest developers I've ever known and became my constant companion on my Mac Plus."

That letter sits among other recollections that were sent to Halls. Decades on from using the software, people remembered the character well enough to write to its maker.

Making a Three-Dimensional Head Feel Alive

A three-dimensional head belonged to the later Windows Moose. Its ears moved and its eyes looked around. Eyebrows and mouth gave it more expressive scope than the original pixels could manage.

Repeating the same animation clips would have made those movements predictable. Algorithms generating its motions gave the Windows character procedural animation instead.

Halls's account of the animation describes controlled randomness. A movement could stop near a target rather than at exactly the same spot each time. The pause before the next movement varied, and its speed varied too. Smooth acceleration and deceleration kept the result from looking mechanical.

A different platform meant a different lip-sync mechanism. Timing information from Microsoft's Speech API told the Windows character which mouth shapes to use. Recreating the old Macintalk memory-observation technique was unnecessary.

Something Worth Saying

Plugins gave the Moose other things to talk about, while jokes remained part of it. To reminders and notifications were added news and weather.

A June 2016 update described a working Pushover connection, IFTTT supplying the events for the Moose to announce. One example gave warning of approaching rain. In another, a phone entering a home geofence triggered a spoken "Welcome home" on the user's PC.

Checking a screen for those messages was unnecessary. They could be heard.

A voice with a face

Forty years stand between the original Moose's 1986 appearance and today's enthusiasm for AI assistants. What Halls actually built would be missed by calling it an early version of today's conversational AI. Speech synthesis and carefully written animation code sufficed to make a joke appear to come from an antlered companion.

For Mike Ellis, that companion belonged on his Mac Plus. The later Windows project could make it the voice announcing rain or greeting someone arriving home.

Looking at the early screenshots is still worthwhile. A face made of pixels, plainly, yet a former user remembered a companion. Quite a result, for a moose that told jokes.