"The premise of Toy Talk as a company is, we want to explore entertainment as conversation. We’re interested in knowing the kind of stories that can be told in conversation."
That’s Oren Jacob. He’s the CEO of a new startup called Toy Talk, but before that, he wrote the Pixar software behind movies like Toy Story and Finding Nemo. And like everyone from Toy Talk, he’s very passionate about a new, old medium in storytelling: conversation.
"I believe we’ll discover there are emotions we can communicate, arcs of stories, and characters to be realized that can only be told specifically through conversation," he says. And that’s not a throwaway quote—he says it in a way that you can tell he really means it.
Toy Talk’s first release is a free iPad app for kids called The Winston Show, a game centered around the character of Winston, a yellow alien-like television host with an energetic British accent, and his quiet, orange ball counterpart, Ellington. What you’ll notice immediately about Winston is just how much he likes to talk—in fact, the app features over five hours of original voice acting, and the studio can record more all the time.
Across a variety of minigames, you interact with Winston not by tapping, like most iPad apps, but by talking to him—holding the mic button to answer a multiple-choice quiz question, or saying the which fork you’d like Winston to take you down on a Choose-Your-Own-Adventure story. But where the app really pushes conversational limits is in Fireside Chats, where you and Winston talk about anything and everything he might bring up.
Winston has asked me everything from what’s my second-favorite sport to why I don’t live in an igloo. And when all goes well, a shorter conversation might go something like this:
Winston: What’s your favorite kind of ice cream?
Me: Rocky road.
Winston: Well, it wasn’t hard for you to get to that answer! You didn’t have to take a rocky road to get to that one! Rocky road ice cream is astoundingly good.
I can understand how that conversation would feel gratifying to a child (because even as a grown man, I appreciated that he understood me and patted me on the head for my ice cream choice). But setting up all the pieces to get this moment just right—to make the software not just understand me but give me a dynamic pun response—takes a combination of creativity, psychology, and science.
"I play Family Feud in my head, where we control the question," explains lead writer Scott Ganz (who in the past, has written for The Muppets), "Therefore, when we ask it, we can think out what are the high-percentage answers we can get. Survey says!"
Speech recognition software caught me saying rocky road. It scanned Toy Talk’s cloud for a rocky-road-specific response that Ganz’s team had penned and recorded. In this case, because there was one, it beamed me Winston’s rocky road pun. But what’s particularly fascinating about this cloud-scripted setup is that it allows Toy Talk’s team to enhance and alter the story based upon audience response.
"We had a question, ‘What was the last thing you ate?’" Ganz tells me. "We had a very wide swath of breakfast foods and sandwiches. Then we started hearing an answer we didn’t anticipate, goldfish—being Goldfish crackers. So we went in, wrote a Goldfish answer and got it into the app."
It’s a simple example for sure, but you can imagine the potential here: Winston doesn’t need to have a knowledge set frozen in time or stuck using anachronistic diction. If his audience is interested in something, he will eventually learn about it. Because if the game’s writers ask for it, they can be greeted each morning with a popular list of words for which Winston had no answer.
But just as fascinating is how these dialogs are dealing with the words they don’t know. Often, Winston will talk around a response he doesn’t understand—offering affirmative vagaries like "I see where you’re coming from and see we’re on the same page!" Once, though, he just gave me a "WTF are you talking about?" look. Winston sat there for a moment. And then we moved on.
It was a very strangely satisfying few seconds—and something I experienced only because the writing team chooses to embrace inevitable miscommunications rather than avoiding them.
"To a large degree, one way a character is communicated is how they respond under stress. In this case, how does Winston respond to an answer they didn't expect?" Jacob explains.
"Having a character is a big advantage to us," adds Ganz. "When we run into those situations that we get an unexpected answer, Winston can express himself and pull the conversation more to himself, while Siri would just say, ‘Let me Google that answer for you!’"
And in that regard, even though voice recognition is one of those technologies perpetually stuck at 90% accuracy, Toy Talk believes their creativity can be agnostic of technological perfection. So long as they continue to leverage the quirkiness of character and clever conversational glue as needed, they can design around the technological potholes and create a new entertainment medium in the process.