Dr. Halls's Talking Moose: An Assembly-Language Original That People Still Remember

In 1986, a black-and-white moose appeared on Macintosh screens and told jokes aloud, his mouth moving in time with the words. Behind him stood a Canadian physician, Dr. Steven B. Halls. He programmed it all himself in assembly language.

The character became familiar to Macintosh owners across North America and Europe. People knew the Talking Moose. Some of them still write to its creator, asking whether he will make a new version.

Few desktop diversions enjoy such a long afterlife.

Antlers, pixels and a mouth that talks

The original Talking Moose had no need for a realistic human face. His antlers made him recognizable on a small, black-and-white screen. Halls was fond of moose, and the outline worked within the limited pixels.

Making him speak convincingly was the harder problem. A moving mouth had to track the speech instead of merely opening and closing during playback.

Halls solved it by observing the Macintosh speech engine, Macintalk, in memory. The Moose could determine what sound was being played and choose a corresponding lip shape. The technique worked because early Macintosh software could inspect memory across applications much more freely than modern systems allow.

The Moose was a watcher, not a modifier, of the speech engine. Code written with care turned speech information into visible movement, so a synthetic voice had a face.

The original character could also comment on window and document names in familiar programs.

A magazine award, and a greeting to remember

The Moose had gained color by around 1989. Halls's pictures of the color version include scans of an image produced with an early Macintosh color printer. The printing dots can still be seen.

The recognition went further than users enjoying the jokes. The software won MacUser magazine's "MacUser Editors Choice Award" for "Best Desktop Diversion" in 1991.

One reception stayed with Halls. At an Apple Developers conference during those early Macintosh years, fellow developers greeted him as "a God". They thought he had coded something impossible.

That was their reaction, and it says something about the surprise of seeing the Moose work. It may seem ordinary now for a face to match a computer-generated voice. On the early Macintosh, Halls made it happen with assembly code and an understanding of the speech engine's memory.

The award and the greeting show two sides of the same program: people enjoyed having it around, while programmers appreciated what it took to build.

Career, children and the requests that kept coming

As the years went on, Halls gave his attention to radiology and family life with children. The Moose no longer sat at the center of everything he did.

But requests for a new version kept arriving sporadically. They still do, even in 2026.

It is one thing for software to be well known while people are using it. Remembering it years later and taking the trouble to contact its author is another matter. Long after those first Macintosh screens, the requests have kept the old character present in Halls's life.

The Moose on Windows, redesigned

In the later Windows project, the character got a three-dimensional head. Its ears could move, and its eyes could look around. The lip synchronization used timing information from Microsoft's Speech API, a different method from watching Macintalk's memory.

Movement changed too. Rather than replaying identical animation clips, the Moose used procedural animation. Algorithms varied where a movement ended and how quickly it happened. Pauses varied as well, with smooth acceleration and deceleration helping the motions look natural.

Plugins widened what the character could say. News and weather arrived alongside jokes, and reminders and notifications gave practical reasons to hear from him.

In a June 2016 development update, Halls reported a working Pushover and IFTTT connection. The Moose could announce approaching rain, or say "Welcome home" when a phone entered a home geofence. By 2019, alpha testers were reporting their experiences with the Windows software, and installer instructions appeared in four languages.

None of it makes the original achievement a story about conversational AI. The Macintosh Moose spoke through programmed behavior and speech synthesis. His mouth moved because Halls had figured out how to match it to the sounds.

A magazine judged it the best desktop diversion. Fellow developers thought the coding was impossible. Decades later, requests for another Moose still reach the physician who proved it could be done.