Newborn Town’s social gaming platform TopTop has introduced Slap Showdown, an AI-powered 3D multiplayer mini-game developed with Viggle, bringing physics-native world model technology into interactive entertainment. The game allows users to turn personal photos into playable 3D characters, combining generative AI, motion simulation and multiplayer gameplay. The launch offers an early example of how world models could move beyond video generation into interactive applications, although their long-term impact on player retention and commercial performance remains to be established.
Viggle Brings Physics-Based World Models to Social Gaming
World models are beginning to find applications beyond AI research and video generation. Newborn Town’s TopTop platform is using the technology in Slap Showdown, a multiplayer mini-game that lets users create personalized 3D characters from photos and bring them into interactive battles.
Developed in collaboration with Viggle, the game is powered by the company’s Joint Space-Time (JST) world model. Players upload a photo and can generate a customized character in approximately 30 seconds, according to the companies. The resulting avatars can resemble anime characters, cyberpunk figures or stylized toy-like models, which players then control through taps, presses and dragging gestures.
The distinction from conventional AI avatar generators is that the generated characters are designed to participate in gameplay, with physics-based movements responding to player inputs. This connects generative AI with interactive simulation rather than limiting it to producing a static image or rendered video.
From AI-generated characters to interactive experiences
Launched in September, Slap Showdown has attracted users in Japan, Spanish-speaking communities in North America and the Middle East, according to the announcement. TopTop reports a 97% match completion rate and says the game-start rate has exceeded the platform-wide average.
The company also cited individual players who completed hundreds of matches over relatively short periods. One player in Japan reportedly played 1,041 matches over 22 days, while another in Eastern Europe completed 735 matches in 11 days.
These figures suggest that some players are engaging frequently with the game, but they are company-reported metrics. Without details about the total player base, retention cohorts, acquisition costs or monetization, they do not establish whether the game is improving platform-wide engagement or generating sustainable revenue.
For TopTop, the broader opportunity lies in connecting user-generated content (UGC) with social interaction. Personalized characters give players a way to express their identity while creating material to share with friends. If the experience encourages users to return, invite others or explore additional mini-games, AI-generated assets could become a recurring engagement mechanism rather than a one-time novelty.
TopTop says future releases could extend AI customization to game environments and introduce new mechanics into existing titles. That would broaden the role of generative AI from avatar creation to more dynamic game content.
Viggle’s JST architecture targets 3D structure and motion
The underlying technology comes from Viggle, an AI company founded in 2022 that focuses on physics-native world foundation models. The company previously gained attention with Meme Maker, a tool for animating static characters and generating videos from specified movements.
Viggle raised $19 million in a Series A funding round led by Andreessen Horowitz in August 2024, according to the announcement. Its longer-term ambition is to extend its world model technology into gaming, social applications and other systems that need to represent movement and physical interactions.
On October 19, Viggle unveiled JST-2, its next-generation model. The company describes it as a step toward models that represent three-dimensional structures and physical dynamics directly, instead of relying primarily on patterns inferred from pixels.
According to Viggle, JST-2 uses a representation built from spatial units with attributes such as position, material, color and mass, alongside information about movement and responses to forces. The model is designed to generate characters, motion and scenes within a unified architecture.
Viggle also claims that JST-2 can generate a 3D scene in seconds for less than one US cent in inference costs. The company says generated content can subsequently be rendered locally without repeatedly invoking the foundation model for every frame. Actual costs will depend on the workload, deployment setup and rendering requirements, and independent performance benchmarks were not supplied in the announcement.
World models face a broader infrastructure test
The development reflects a wider shift in AI research toward systems that model environments and their dynamics, rather than generating content alone. Large language models generate and process language, while world models aim to represent how objects and environments behave over time. That capability could support applications ranging from interactive games to simulation and robotics.
The commercial landscape remains competitive. NVIDIA provides simulation and robotics infrastructure through its Omniverse platform, while Google DeepMind has researched systems that generate interactive environments. These efforts differ in architecture and intended use, but illustrate the wider industry interest in AI systems that connect digital representations with spatial reasoning and physical behavior.
For Viggle, the challenge is to demonstrate that its physics-native approach offers practical advantages in quality, controllability, latency and cost. For TopTop, the priority is more immediate: making AI-powered interactions sufficiently entertaining and reliable to sustain player engagement.
Viggle says its planned JST-3 model will aim to function as a broader world simulator, potentially supporting game engines and virtual training environments for robotics and other physical AI applications. That remains a future ambition rather than a demonstrated capability.
Newborn Town has also invested in AI-related areas including AI companions, agents and interactive gaming, according to the announcement. The TopTop partnership gives the company a practical setting in which to test world model technology with consumers.
The significance of Slap Showdown is therefore less about proving that world models are ready to transform gaming as a whole and more about demonstrating a potential path from 3D generation to playable experiences. Whether that approach scales will depend on technical performance, user demand and evidence of lasting business value.
Market Landscape
World models are an emerging part of the generative AI landscape. Their goal is to represent environments, objects and physical dynamics in ways that can support prediction, simulation and interaction. Gaming provides a useful testing ground because generated content must respond to user input rather than simply look convincing.
Viggle’s JST approach competes within a broader ecosystem that includes AI research into interactive video, 3D generation, simulation and robotics. NVIDIA’s Omniverse illustrates the role of simulation infrastructure, while Google DeepMind’s work highlights interest in generative interactive environments. These approaches should not be treated as technically equivalent.
For enterprise AI developers, key evaluation criteria include generation latency, inference costs, consistency of motion, controllability, safety and integration with existing rendering engines. Social gaming adds another measure: whether personalization increases repeat use without making the experience slow or difficult to access.
No directly comparable independent benchmark was provided for JST-2. Its reported sub-cent scene-generation cost and seconds-long generation time should therefore be treated as vendor claims rather than established industry benchmarks.
Top Insights
- Playable AI characters: TopTop combines photo-based 3D character generation with physics-based motion, taking generative AI beyond static avatars and prerecorded video.
- Early engagement signals: The company reports a 97% match completion rate, but broader retention and monetization data are needed to assess commercial impact.
- Physics-native architecture: Viggle says JST-2 represents spatial and physical properties directly, targeting more controllable 3D characters, motion and scenes.
- Cost claims need validation: Viggle reports sub-cent scene generation, although independent benchmarks and workload-specific cost comparisons were not disclosed.
- Broader AI ambitions: Viggle’s planned JST-3 model targets simulation and physical AI applications, but those capabilities remain under development.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
