Newsletter
How Snickers Cloned José Mourinho and Won 2 Grand Clios
The coach tells it like it is in AI-powered campaign
Who better to roast your friends than José Mourinho?
Snickers delighted football fans the world over by creating an “Own Goals”-inspired AI tool. It allows users to tell the iconic coach about some dumbass mistake made by their bestie, brother—or anyone in their lives and receive a personalized video of Mourinho offering advice.
The project scored a pair of Grands at last night’s Clio Sports Awards in NYC.
T&Pm worked with the production company Helo, as well as ElevenLabs, Synclabs and Open AI GPT 4.0 to create the AI-powered version of Mourinho. Meta helped integrate the clone into WhatsApp.
Here, T&Pm group creative creative director Jim Stump takes us inside the campaign.
MUSE: What makes this AI-led work so groundbreaking?
Jim Stump: Until now, we’ve seen many authorized—and dubious non-authorized—AI versions of celebs. Bruce Willis speaking Russian in a commercial, Harrison Ford de-aged via AI, fake Gareth Southgate England team press conferences during the Euros. But the films are always linear. We wanted to let football fans actually affect what José Mourinho might say and generate their own funny put-down they’d want to send to a friend who’s messed up and needs a Snickers.
Of course, building strong enough safety elements to avoid José saying undesirable things was hugely important. But this, hopefully, is a step in a new direction for ways talent can allow their image to interact with fans.
I imagine the level of teamwork and communication required to make this was incredible.
Absolutely. We developed the entire project from beginning to end in 15 weeks, and this included a three-week R&D phase to establish initial plausibility. But with new AI audio and lip-sync technologies coming out every week, the guys at Helo [the production company] were building the ship once we’d set sail.
Such tight deadlines meant we had to work extremely closely with them, key client stakeholders and Mars legal, guidance and governance. Kyle Ruddick, Helo’s lead AI experience director, is based in L.A., so almost daily Zoom standups with him and the Helo team meant we could keep everyone up to speed.
Once we were happy with the quality of AI José, how he looked, sounded and the humor of his responses, attention fell to the level of safety. We were all “red teaming,” trying to hack and break the build—getting it to say things it should’t. We lost whole afternoons discussing the moral implications of allowing José to say certain things and not others. Even innocent words can be used with malicious intent. We had to flesh out a glossary of over 5,000 off-limit words and phrases.
When you approached José, was he into the concept?
José and his team were very interested and supportive from the beginning, thankfully. They’re aware of what AI is enabling in the realm of talent rights. For him, it entailed a three-hour film shoot. Naturally, they wanted to ensure the visual and audio capture was as true to the real José as possible and to stay close to the safety elements to protect him. We regularly checked in with key updates as we went along.
What was the one big takeaway from this project?
That there’s a world of difference between making AI films and making a tool for people to create their own AI films. Most people who made one video went on to generate one or two more. That’s huge engagement time with the brand and testament to just how humorous and surprising generative AI can be, even when you put strict parameters in place.
