In the video, a crowd roars at a packed summer music festival. Beats start to play through the speakers, and finally, a performer steps onto the stage: The Joker. Dressed in a red suit, green hair and distinctive face paint, he pumps his fists, dances around the stage, and jumps down the runway to get even closer to the sea of fans. As he begins to rap, the Joker bends his knees and leaps off the ground, bouncing up and down before spinning 360 degrees on one foot. It looks effortless, but if you tried to pull this off in real life, you'd fall on your face. The Joker has never been this cool.
Then there's NBA All-Star Joel Embiid strutting out from backstage, greeting the crowd, and performing the same dance moves. Then there's “Curb Your Enthusiasm” star Larry David. But something about each of these scenes is just a little off. Whether it's the Joker, Joel Embiid, or Larry David, the performer's bodies are shaking, but their expressions stay perfectly still.
Of course, this is all AI-generated, thanks to a company called Viggle.
The original video shows rapper Lil Yachty taking the stage at the 2021 Summer Smash Festival, and according to the title of the YouTube video, which has more than 6.5 million views, the entrance is the “Hardest Exit Ever.” In April, this became a trending meme format, with people inserting their favorite celebrities and favorite villains, like Sam Bankman Freed, into the video of Lil Yachty taking the stage.
AI services that turn text into videos are getting scary good, but you can't expect Sora to accurately understand what you mean when you type in “Sam Bankman-Fried as Lil Yachty in 2021's biggest summer hits.” Viggle works differently.
In Viggle's Discord server, users upload a video and a photo of themselves doing some kind of move (often a TikTok dance). Viggle then creates a video of that person mimicking the moves from the video. These videos are obviously not real, but they're still funny. But after the Lil Yachty meme went viral, Viggle became popular, and it's not slowing down.
“We're focused on building what we call a controllable video generation model,” Viggle founder Hang Chu told TechCrunch. “When generating content, we want to have precise control over how characters move and how a scene looks. But current tools focus only on text-to-video conversion, and text alone is not enough to specify all the visual nuances.”
Chu says there are two main types of users on Viggle: those who create memes and those who use the product as a tool in their game design and VFX production process.
“For example, our team of animation engineers can take some concept designs and quickly turn them into rough animation assets,” Chu says. “The idea is to see what it will look and feel like with rough sketches of the final plan. Typically, this would take days or even weeks to set up manually, but with Viggle, it can be done automatically, essentially instantly. This significantly reduces tedious and repetitive modeling work.”
In March, Viggle's Discord had a few thousand members. By mid-May, that number had reached 1.8 million, and just days before June, Viggle's servers had more than 3 million members — more than the servers for games like Valorant and Genshin Impact combined.
Viggle's growth shows no signs of slowing down, except that high demand for generated videos means wait times are a bit too long for impatient users. But because Viggle is Discord-centric, Discord's development team works directly with Viggle, guiding the two-year-old startup's rapid growth.
Fortunately for Viggle, Discord has been in this situation before: MidJourney, which also runs on Discord, has 20.3 million members on its servers, making it the largest single community on the platform. Overall, Discord has around 200 million monthly users.
“Nobody's ready for that kind of growth, so we start working with them at that virality stage, because they're not ready,” Ben Shanken, Discord's VP of product, told TechCrunch. “We have to be ready. The majority of messages being sent today are from Viggle and MidJourney, and a lot of the consumption and usage on Discord is actually generative AI.”
For startups like Viggle and MidJourney, building an app on Discord means they don't have to build an entire platform for their users. Instead, they can host their app on a platform that already has tech-savvy users and has content moderation tools built in. For Viggle, which has just 15 employees, Discord's support is essential.
“We can focus on building our model as a backend service while Discord can leverage their infrastructure on the front end, essentially allowing us to iterate faster,” Chu said.
Prior to joining Viggle, Chu worked as an AI researcher at 3D tools giant Autodesk and has also conducted research at companies such as Facebook, Nvidia and Google.
For Discord, accidentally operating as a SaaS company for AI startups could be costly. On the one hand, these apps will bring new users to Discord, which will have a positive impact on user metrics. But hosting so many videos when other users on the platform are streaming live video games, video chats, and voice calls could be technically challenging and costly. But without a platform like Discord, these startups may not be able to grow at the same rate.
“Scaling is not easy for any type of company, but Discord is built for that scale and we're able to help companies successfully absorb that scale,” Shanken said.
These companies can adopt Discord's content guidelines and even use content moderation apps, but making sure 3 million users are following the rules is always going to be a challenge. Even Lil Yachty's exit meme technically violates Viggle's rules, which encourage users not to generate images of real people, including celebrities, without their consent.
For now, Viggle's saving grace may be that its output isn't 100% realistic yet — the technology is truly impressive, but we're well aware — and while that awkward Joker animation is definitely not the real thing, it's certainly entertaining.