Netflix has unveiled a talking, human-like interface named “Max,” who greets subscribers with a game-show host’s exuberance and invites them to play irreverent games, such as “Max’s Mystery Call” and “Celebrity Mood Ring.” It seeks to inject humor into the analytical process of offering viewing suggestions — and to further differentiate the service from competitors such as Amazon.com or Hulu.
Subscribers who select Max from Netflix’s main menu launch one of four game-like experiences designed to refine the suggestions. In “Celebrity Mood Ring,” Max reviews a user’s viewing history and selects two actors who represent distinctly different genres: say, action star Jackie Chan and Academy Award-winning dramatic actress Kate Winslet. For the person who selects the martial arts master, Max might proffer the 1991 action adventure film “Operation Condor.”
“People have wide tastes,” said Todd Yellin, Netflix’s vice president of product innovation. “When they talk to Max, he finds out what they’re in the mood for.”
Max is the technological love child of Netflix’s sophisticated algorithms, which offer movie and TV suggestions based on an individual’s viewing preferences, and the work of Jellyvision, the Chicago company (best-known for its “You Don’t Know Jack” interactive trivia games) that specializes in bringing character to inanimate technologies.
“Our job, as the peanut butter to Netflix’s chocolate, was to bring all of the numbers of the recommendation algorithm to life, layering on top of it a voice, a personality,” said Jellyvision Founder and Chief Executive Harry Gottlieb. “A huge part of the effort was about creating a character, with Max, who is not your typical movie reviewer on ‘Good Morning America.’ It’s more like a friend talking to you.”
Netflix and Jellyvision began working together in 2006 to develop a ratings recommendation game, which suggested new titles based on ones the subscriber selected online. Netflix tested the game in 2007 but decided to shelve the concept, which had limited appeal at a time when most subscribers received their movies on DVDs mailed in little red envelopes.
However, consumer acceptance of speech recognition technology began to increase with Apple Inc.’s 2001 introduction of the Siri personal assistant, which popularized the sci-fi notion of people speaking commands to their gadgets. That same year, Microsoft Corp. brought voice recognition to its Xbox 360 game console through its Kinect sensor. Google, meanwhile, has been dabbling in voice search since it introduced the Goog-411 telephone directory in 2007, and has brought the same features to mobile devices and the desktop.
“Before, these were boxes that did jobs for you. Now they are really entities with a personality. Mobile devices and tablets are becoming so popular, these kinds of attributes are becoming more and more important,” said Abeer Alwan, a professor of electrical engineering at UCLA. “I think there is more acceptance and more admiration for what these machines can do.”
Evolving consumer attitudes set the stage for Netflix and Jellyvision to renew their work animating the service’s recommendations. Max was born about a year and a half ago.
Writers scripted thousands of lines of dialogue – sentence fragments that could be stitched together dynamically, so Max appears to be conversing with the Netflix subscriber and reacting to what’s happening on the TV screen.
“Max seems aware of what you did last session with Netflix, or the time of day, saying, ‘Here’s a lunchtime suggestion,’ ” Yellin said. “I think part of the magic is going to be using creativity and technology to expand these branches and make them feel more sentient.”
Netflix and Jellyvision auditioned about 150 voices for the part of Max, and cast six female voices – and one male voice – to test with users.
“What’s interesting is this is not your typical acting work – it’s not even typical voice-over work,” Gottlieb said. “If you’re in a play, you get to be in character for a full hour and a half. In a movie, you’ll do scenes – maybe perform a few lines of dialogue. Here, you might be performing parts of a sentence four different ways, and it will be concatenated [or arranged in a string that forms a line of dialogue]. Frankly, not every actor can do this.”
Yellin said Netflix and Jellyvision initially believed that subscribers would be more receptive to a woman, named Maxine, because of Siri and the female voices heard calmly providing turn-by-turn directions on most navigation systems. Initial tests seem to support this hunch.
“Maxine hands down was crushing Max,” Yellin said. “People said she was friendlier, smarter, more approachable.”
But in tests with about 50,000 Netflix users, Max emerged as the voice subscribers preferred -- although Yellin struggled to explain why.