Inworld AI Debuts Metaverse Character Builder
Virtual being startup Inworld AI has launched its Inworld Studio platform for designing interactive characters for the metaverse and other digital worlds as a beta for developers. Inworld Studio generates its interactive AI in just a few minutes from the stat bars and text forms seen in the menu above. The resulting AI personality will then respond and evolve based on their conversations and experiences in the virtual environments.
Inworld Studio’s engine constructs virtual beings out of what the user writes in the text fields, supplemented by the slider bars. That gives creators a huge amount of flexibility to define their AI personality to the smallest detail. They can fill in a whole history for the virtual character; their history, family, interests, experience, and knowledge. Inworld offers some general knowledge databases, so the user doesn’t have to write out an entire 12-year curriculum. They can also connect the AI to other knowledge sources, such as a company’s background for a corporate representative. Once the personality is set, the AI can be linked to a look and voice from Inworld’s library of licensed avatars and embedded as an API in its new virtual home in the metaverse or wherever the designers want to set it up.
“The platform is super easy; there’s no coding. You can fill in the text forms with whatever you want,” Inworld AI chief product officer Kaylan Gibbs told Voicebot in an interview. Gibbs headed Deepmind applied generative language model team before joining Inworld.
Gibbs demonstrated Inworld Studio for Voicebot by producing a virtual Spider-Man (in his secret identity of Peter Parker). He wrote a series of statements based on suggestions as to the superhero’s interests, personality, and goals and plugged it into a standard example of a male character. Within a few minutes after finishing, Gibbs began chatting with the virtual Spider-Man, who was happy to chat about his Aunt May and his love of photography, helping people, and making jokes. Virtual Spider-Man is not quite ready to replace Tom Holland and lacks the polish of the more fleshed-out example characters in both personality and voice. That said, the AI character stood out as remarkably usable for such a limited amount of time and effort. An entire developer team spending a couple of weeks with the platform could likely come up with a multi-purpose brand avatar ready to launch.
The fluid personalities of Inworld Studio’s virtual characters don’t vanish when the API is embedded within a platform. The AI is designed to mimic how humans think and react to the world around them. That means the AI will react to its environment based on its personality, being happy to see its favorite animal or annoyed by whatever was described as a pet peeve. The same goes for interacting with actual humans. What people say and the emotion underlying the words will provoke a response. Give an AI a favorite sports team and tell it the team won and it will look and sound happier, but tell it you were kidding and it will sound much sadder. The AI won’t immediately rest to neutral but will remember what happened, which will influence what and how it speaks in the future. Inworld’s ambition is for AI personalities to change and learn from real-time events.
“We want it to feel like you are talking to another human who can understand your words and the feelings [behind] your words,” Gibbs said. “You can fine-tune [the character] as much as you want, but it also changes from what happens in your conversations and in the virtual space.”
Inworld was in stealth until late last year but has now raised nearly $20 million in venture funding. A $10 million strategic funding round in March has given the startup some space to continue improving Inworld Studio during the beta test. Speed and functionality are both crucial as virtual beings continue their rise as a central pillar for conversational AI. While the metaverse as a concept and market is in some ways undefined, and demand is accelerating for increasingly realistic digital characters to interact with everywhere, from video games to virtual corporate campuses.
Digital award shows, basketball games, and other experiments are part of why Inworld and so many other startups with a toe in the market are raking in cash. Inworld’s $10 million news came in the middle of the flood, including Hour One’s recent $20 million, a $21.5 million funding round by Neosapience, and a tidy $70 million raise for Soul Machines, to give a few examples. Inworld boasts a few ringers that give it an edge, however. CEO Ilya Gelfenbeyn founded the startup that became Dialogflow after Google acquired it. And their advisors include Academy Award-winning designer John Gaeta, whose ground-breaking work made The Matrix and much of modern cinema possible.
“Inworld AI is powerfully pushing forward the types of experiences possible within interactive worlds,” Gaeta said. “by enabling creators to develop intelligent characters who naturally interact, emotionally behave, intuitively learn and individually express who they are in ways we’ve never seen.”