Graphic design of cutout mouths talking and layered over each other
An arrow pointing leftHome

This is what speech-improvement AI has to say about five celebrities with infamous voices

  • Mike Pearl

In which we feed celebrity speeches into an AI designed to offer automated feedback to see how they do.

Yoodli is a tech startup based in Seattle that is building an AI tool that analyzes speech throughout the day, in various contexts, and provides users with automated feedback. Over time, users will hopefully apply that feedback, and improve their confidence and overall communication abilities. The young AI company, which I’ve previously dug into here, is funded by Madrona Venture Group and the Paul Allen Institute for Artificial Intelligence (also a key funder of the publication you’re currently reading!).

When I spoke to co-founder and CEO Varun Puri, he painted a picture of a piece of tech that might involve wearables, or perhaps an always-on microphone, along with some heavy-duty processing power cranking away all day — ambitious stuff. At the moment, however, there’s a very useful tool already available at the Yoodli website. It’s clearly labeled “Yoodli (Beta),” but I didn’t let that status stop me from using, and possibly abusing it.

I plugged a bunch of celebrities and public figures into Yoodli, and had the software lambast them for their filler words and hedging, and offer encouraging comparisons to, well, other celebrities with similar speaking problems.

Overall, Yoodli is a pretty generous robot. As a former English teacher myself, I tended to dole out more demerits than gold stars (I never said I was a former good English teacher), whereas Yoodli is mostly encouraging, even when a speech is pretty rough. But Yoodli undeniably also zeroes in on problem areas in the same ways a human listener might. And from time to time, Yoodli can be a little harsh.

For each type of evaluation, users receive a green thumb up, or a red light bulb from Yoodli. The maximum score is 10 green thumbs.

1. Gilbert Gottfried: 9/10 green thumbs

Gilbert casts light on the inspiration behind Melville's famous book.

I’m a massive fan of the late comedian Gilbert Gottfried, and I was delighted to find that Yoodli is too. It applauded his “relaxed” speaking pace, and congratulated him on his 96% “articulation” — Yoodli’s measure of intelligibility. I thought it might knock Gilbert for “loudness,” but it seemed to find him “average,” although its graph did show that he got significantly quieter and then louder, while remaining within the acceptable range.

Gilbert Gottfried's "loudness" during speech.

It did somewhat unfairly knock him for “non-inclusiveness” when he used the word “Dick,” but the Dick in question was of the “Moby” variety.

2. Fran Drescher: 9/10 green thumbs

Fran Drescher discusses what growing up with such a distinctive voice did for her career.

Neither Gilbert Gottfried nor Fran Drescher triggered any alarms in Yoodli’s analysis software about being too loud or unclear, so maybe we humans should reconsider those thoughtless criticisms before we hurl them at these beloved entertainers.

Drescher did get dinged for hedging too much. If 3% or more of your speech is made up of hedging phrases, Yoodli will switch you from a green thumbs-up to a red light bulb in that category. She used “kind of” and “I think that” once each, and “so” a worrying five times to start sentences. One of her more troubled sentences was, “So I was in high school wanting to break into the business and that kind of, uh, young person was, you know, being seen on TV a lot.” Tsk, tsk.

3. Julie Andrews: 8/10 green thumbs

Julie Andrews accepts her AFI Life Achievement Award.

I figured Julie Andrews was the anti-Fran Drescher, since she’s also known for playing nannies, but her voice is famously soothing, authoritative, and downright musical. Did Mary Poppins perform better than TV’s The Nanny? No! In fact she fared worse.

Not only did Andrews tie Drescher for hedging, she also got a slap on the wrist for repetition. This seems to be a choice Andrews makes, however, rather than a mistake. “They are, they are the unsung heroes of moviemaking,” she says in her speech, getting more emphatic on the second “they are.” Is this type of thing poor communication? It’s not for me to say.

4. Alton Brown: 4/10 green thumbs

Precise Advice with Alton Brown

There’s no easy way to say this: Yoodli absolutely clobbered TV chef Alton Brown. He used an astonishing 6% filler words — mostly “uh” — and 3% hedging phrases, and these phrases crept into his “keywords” box, meaning he repeated them so much that things like “actually” and “uh” became the subject of his speech. At 200 words per minute, he was too fast. Perhaps worst of all, not only was he a word repeater, but many of the words he repeated were themselves filler.

Yoodli's evaluation of Alton Brown's speech. It's not looking pretty.

Brown did get dinged unfairly for the word “fat.” He wasn’t commenting on someone’s body; he was talking about the cooking element.

Overall, however, Mr. Brown needs considerable practice if he’s ever going to make it as an entertainer. According to Yoodli, anyway.

5. James Earl Jones: 9/10 green thumbs

James Earl Jones accepts the Lifetime Achievement in the Theatre Award during the Creative Arts Awards portion of the 2017 Tony Awards ceremony.

I included acting legend James Earl Jones because I’d heard the famous anecdote about his heroic struggle to speak without a stutter. That makes it all the more satisfying to see that Yoodli passed Mr. Jones with flying colors. His only filler word was “okay,” and his only hedge was “so.” Interestingly, the only admonition from Yoodli was for Jones to speak more quickly, something I haven’t seen very often in all my use of the app. It also offers the recommendation, “Vary your pace to keep your audience engaged.”

But would you tell Darth Vader to talk faster and vary his pace because people aren’t engaging with him? I think I know what would happen if I did that.