AWS for Games Blog
Kinetix taps AWS to bring generative AI tools directly to players
With social interaction central to the gaming experience, demand for user-generated content (UGC) across games is exploding, from Roblox to Fortnite and beyond. Just as no two humans think, talk, or move the same way, many video game development studios and players are of the mindset that digital characters shouldn’t either. This is where emotes (custom 3D animations) are changing the game, allowing players to establish more distinct identities when interacting with others in-game. Taking emote capabilities to the next level, Paris-based artificial intelligence (AI) technology startup Kinetix developed a new generative AI-enabled tool that lets players create emotes that reflect their personalities, emotions, and tastes.
Four years in the making, the solution allows game publishers and developers to put gamers, and community developers, or modders, in the driver’s seat with generative AI. Once a developer integrates Kinetix technology into a game, players can generate custom emotes based on internet or self-shot videos and apply them to avatars, with protective content moderation layers built in. Already used by 25 percent of alpha testers exploring Krafton’s UGC metaverse OVERDARE, Kinetix’s personalized AI emote creation tool is powered by Amazon Web Services (AWS) technologies.
Focus on the technology
Designed to enhance gaming experiences, Kinetix’s AI emote technology makes it easier for developers and publishers to support custom UGC. A progressive web app (PWA) running on the player’s phone enables them to send videos to the Kinetix AI emote tool. Behind the scenes, the company’s proprietary AI and machine learning (ML) algorithm performs the heavy lifting; it analyzes the video, generates the 3D animation, and applies it to the avatar.
Rooted in computer vision, the solution receives video and analyzes each frame. It identifies the human body and estimates the position of each bone. From there, Kinetix’s technology applies a 3D animation to the skeleton based on pose estimations provided by the frame-by-frame analysis. Retargeting technology then identifies and adapts to differences between the standard Kinetix avatar and the player’s custom avatar, ensuring animations are compatible with any 3D avatar. Kinetix’s AI creates animations for standard-shaped avatars, and then, the animations are adapted to take into account the special morphologies of player avatars.
To prevent explicit, vulgar, or offensive content from making its way into any avatar animation, Kinetix’s team built a video moderation layer into the solution, which uses Amazon Rekognition to analyze video content. If it detects any unwanted movements, the solution changes the animation to make them disappear. Before finalization and sharing, the emote undergoes a secondary screening using proprietary technology.
Enabling user-generated emotes for OVERDARE
Kinetix’s solution is gaining attention for its potential applications in the UGC metaverse, with the technology fully integrated into OVERDARE. Developed by Krafton, the South Korean game studio behind popular battle royale shooter PUBG, OVERDARE is a platform that allows users to create their own games and interactive content with a create-to-earn (C2E) business model. Set to launch by the end of 2024, it runs on servers powered by AWS and is currently in an alpha test phase on Android in Southeast Asia.
Kinetix worked closely with Krafton to customize its AI emote technology integration for OVERDARE, aligning with Krafton’s unique platform goals. With the world of OVERDARE built in Unreal Engine 5, Kinetix was able to leverage its Unreal Engine SDK (Software Development Kit) to ensure a smooth integration before alpha release in May. Players from Thailand, Malaysia, Singapore, the Philippines, and Indonesia have already started experimenting with the UGC platform, 25 percent of whom have created emotes using a feature powered by Kinetix’s technology.
“This is the first time we’re putting generative AI in the hands of gamers, and as a gamer, I’m super excited about that. I can’t wait to see what OVERDARE users build with the technology,” shared Alex Breg, Kinetix Marketing and Business Associate. “Leveraging technology from AWS, we’ve built a solution that enables developers and publishers to give their gamers incredibly intuitive tools for creating amazing UGC in-game. It’s the first step in a long journey with generative AI that will ultimately make for better player experiences within the gaming community, and we’re so proud to have launched the technology with a partner like Krafton.”
Exploring new possibilities for the modding community
As Kinetix has continued to develop its AI emote technology, the modding community has taken notice. For example, modders have incorporated Kinetix technology into custom game modes of a number of popular titles, with the new functionality introduced from generative AI receiving much enthusiasm.
“Historically, UGC has only been possible for modders with development skills or players who can code, but the work we’re doing with AI is set to unlock emote opportunities for players with no coding experience to bring custom animations into their mods,” Breg explained. “What has been accomplished thus far is only the beginning. It will only become easier to create something unique that’s just yours to share with other players as we continue to develop the tech and the community adopts it.”
Scaling for the future
OVERDARE marks the first of many native integrations for Kinetix as UCG demand continues to grow. As more active users create emotes, Kinetix can easily scale its AWS servers to meet demand. Kinetix is also exploring how Amazon SageMaker might help it build, train, and deploy ML models at scale.
Kinetix’s solution is supported by a host of AWS services, including Amazon Elastic Block Store (Amazon EBS), AWS Fargate, Amazon Elastic Compute (Amazon EC2), Amazon Elastic Load Balancing (Amazon ELB), Amazon Elastic Container Service (Amazon ECS), Amazon Relational Database for PostgreSQL (Amazon RDS for Postgre SQL), Amazon S3, Amazon Web Application Firewall (Amazon WAF), Amazon Cognito, Amazon CloudWatch, Amazon OpenSearch Engine, AWS CodeArtifact, AWS Lambda, Amazon CloudFront, AWS Step Functions, Amazon Elastic Container Registry (Amazon ECR), AWS Systems Manager, Amazon DynamoDB, Amazon Elastic File System (EFS), Amazon DocumentDB, Amazon Virtual Private Cloud (Amazon VPC), Amazon Simple Queue Service (Amazon SQS), Amazon Simple Notification Service, and Amazon Simple Email Service (SES).