Despite its growing popularity, generative AI is unlikely to become widely adopted on Discord, experts say, due to various challenges and limitations
A raucous crowd can be seen in the video at a crowded summer music festival. As a rhythmic pattern emerges from the speakers, the eventual performer emerges onto the platform: it is the Joker.
The Joker, decked in his signature face paint, green hair, and red suit, pumps his fist and dances across the stage, hopping down a runway to approach his sea of admirers even further.
The Joker initiates his rap by knee-flexing and propulsion off the ground, bounding up and down before performing a 360-degree turn on one foot.
Despite its seemingly effortless appearance, the execution of that maneuver would render you face-planted. Never has the Joker appeared this suave.
Another video features NBA All-Star Joel Embiid performing the same dance moves while strutting from backstage to greet the audience. Larry David, protagonist of “Curb Your Enthusiasm,” follows.
However, each scene has a slight irregularity: the actors portraying the Joker, Joel Embiid, or Larry David have unsteady bodies, and their facial expressions remain static.
AI-generated everything here, courtesy of the company Viggle.
The initial video depicts Lil Yachty performing at the Summer Smash Festival in 2021; a YouTube video with more than 6.5 million views claims this entrance is “the most arduous exit ever.”
In April, this became a popular meme format, with individuals inserting their preferred celebrities—or antagonists, such as Sam Bankman-Fried—into the Lil Yachty performance video.
Converting text to video, Sora cannot comprehend your query, “Sam Bank man-fried as a little yachty at the 2021 summer smash,” despite frighteningly improving AI offerings. Viggle functions otherwise.
Users upload a photo of themselves and a video of another individual performing some movement (typically a TikTok dance) to Viggle’s Discord server.
Viggle then generates a video of the individual imitating the movements seen in the video. Although it is evident that these videos are not authentic, they continue to amuse.
However, Viggle became popular after the Lil Yachty meme went viral; the excitement has yet to subside.
“Our primary objective is to develop what we refer to as the controllable video generation model,” said Hang Chu, the founder of Viggle.
According to Chu, Viggle is utilized by two primary categories of users: those who create parodies and those who use the application as a resource for game design and visual effects production.
“A group of animation engineers, for instance, could rapidly transform some concept designs into rough animation assets,” Chu explained. “Seeing how they appear and feel in the rough sketch of the final plan is the entire purpose.”
This process typically requires days or weeks to complete manually, but Viggle can accomplish it virtually instantaneously and automatically. This eliminates a substantial amount of laborious and repetitive modeling effort.
Viggle’s Discord had several thousand users in March. As of mid-May, the number of members on Viggle’s server stood at 1.8 million. As June approached, that figure rapidly increased to more than 3 million.
This makes it more significant than the aggregate server capacity of games such as Genshin Impact and Valorant.
Aside from the high demand for video generation, which has caused lengthy wait periods for impatient users, there is no indication that Viggle’s expansion will slow.
However, due to Viggle’s intense reliance on Discord, the developer team of Discord has collaborated closely with the two-year-old startup to guide its rapid expansion.
Discord has, to Viggle’s benefit, been through this situation previously. With 20.3 million members on its server, MidJourney, which also operates on Discord, is the largest singular community on the platform. Discord has approximately 200 million monthly users.
“No one is prepared for that type of growth, so we begin working with them during the virality stage because they are not prepared,” said Ben Shanken, vice president of products for Discord.
“We must be prepared, as a significant portion of the messages being sent at the moment consists of Viggle and MidJourney, and a substantial amount of Discord usage and consumption is generative AI.”
By utilizing Discord to host their applications, entrepreneurs such as MidJourney and Viggle can circumvent the need to develop an entirely new platform for their users.
Instead, their applications are hosted on a platform that possesses an established technological community and incorporates content moderation tools. Because Viggle has only fifteen employees, Discord’s assistance is vital.
“It is possible to iterate faster if we concentrate on developing the model as the backend service and Discord utilizes their infrastructure for the front end,” Chu explained.
Chu was an AI researcher at Autodesk, a 3D tools company, before joining Viggle. Additionally, he conducted research for Facebook, Nvidia, and Google, among others.
Inadvertently serving as a SaaS provider for AI startups could harm Discord. While these applications attract a fresh user base to Discord, they likely contribute positively to user metrics.
However, hosting such a large amount of video can be challenging and expensive technologically, particularly when other platform users are streaming live video games, video chatting, and voice calling. However, these firms’ growth rates might only be impeded by a platform such as Discord.
Shanken stated, “Scaling is difficult for any type of business, but Discord is designed for that type of scale, and we can assist them fairly well.”
Although these businesses can utilize Discord’s content moderation apps and employ its content guidelines, ensuring that three million users behave appropriately will always be challenging.
Technically, even those Lil Yachty walk-out parodies violate Viggle’s policies, which advise users to obtain permission before creating images of real people, including celebrities.
Viggle’s saving grace may be that its output could be more realistic. Even though the technology is genuinely remarkable, we know better. Although that sloppy Joker animation is certainly not actual, it is humorous.
