Tassel at ASU

ASU Graduations are now optimized with AI voices
(Posted by Peter Lafford, May 10, 2025.)

At the graduation, you hear the name over the PA system, “Robert…Louis…Stevenson.”

Was that recorded by a human being? Or was that an AI-cloned voice? You might be surprised by the answer!

One way or the other, Peter Lafford's voice will be heard announcing 9,000+ graduate names at ASU's commencement and convocation events in Tempe, this week. Peter has been announcing names at various ASU graduation and convocation events for 30 years, originally by live-reading the names scribbled on a card by the graduate at the event. Over the years, technology from graduation-coordination company MarchingOrder (now named “Tassel”) has refined the process, by allowing the readers to prepare the names and make notes that would appear on the name card, and more recently, by having the name readers pre-record the names to be played as the graduate crosses the stage. Now, Tassel and AI technology have advanced the process to a new phase.

Tassel assigns names to four readers for ASU’s graduations. This Spring, Tassel assigned 9,000 of ASU’s 21,000 graduate names to Peter’s voice, and assigned the other 12,000 ASU names to the three other readers. Before each reader records the assigned names, Tassel submits all the graduates' names for processing by an AI-clone of the assigned reader’s voice. When the graduates RSVP to attend graduation on the Tassel website, the graduates themselves can approve or reject the AI-produced result. With about an 80% approval rate of the AI pronunciations, about 20% of the names end up in the Reader-Record queue, so Peter was left with about 2,000 names to do as a live recording in his studio a few weeks later; the other readers live-recorded about 2,500 of the 12,000 remaining names. Peter says he fine-tunes the live-recorded pronunciation to be just the way the graduate wants it, after she or he enters the phonetic pronunciation and leaves a recording of the name on the Tassel RSVP website.

While Peter acknowledges that he listens to the AI voice recordings before doing his own live recordings, so he can emulate the deliberate and careful AI voice pronunciations, Peter says he is hard-pressed to distinguish the AI-produced names from his live-recorded names, and is quite impressed by the AI cloned-voice technology. Tassel says they can attain greater accuracy with more names, and provide more name-announcing consistency across the various events at a given institution.

One of the other benefits of the Tassel graduation package is that, at the on-stage podium, when the graduate presents the Tassel QR code on the name card (or at some events, on a cell phone), the graduate’s name will be displayed on the big screen AS the recorded name is played, while the graduate smiles for the mandatory graduation photo. When graduates come from two sides of the stage, this can accomplish 3-second-per-graduate efficiency, getting upwards of 1,200 graduates across the stage in an hour. That can be much appreciated in a hot Mountain America Stadium evening event!

So as you listen to the names at the next graduation event you attend, ask yourself, was that recorded by a human being? Or was that an AI-cloned voice? You might be surprised by the answer!