Particle 6 Group Debuts AI Actor Tilly Norwood
Particle 6 Group has opened public video calls with its generative AI actor Tilly Norwood, highlighting the technical and ethical hurdles facing digital avatars in entertainment.

Particle 6 Group, through its AI division Xicoia, has launched an interactive, British-accented digital character named Tilly Norwood. Unveiled at the Zurich Film Festival, the avatar is designed to act as a protagonist in an upcoming feature-length project titled Misaligned, set in a digital world called the Tillyverse. To promote the project, the studio opened public video calls with the character between September 16 and September 19. After a free two-minute trial, users paid $14 for an extra 15 minutes or $25 for an additional half hour, with proceeds reportedly going to charity.
Built using several publicly available third-party AI tools, the interviewable version of Norwood runs on a system distinct from the core animation model. The system features visual appraisal capabilities, allowing the avatar to comment on a user's clothing. However, tests by journalists revealed significant conversational limitations. The chatbot struggled to provide concrete details about its film, such as an approximate release date, and frequently redirected questions back to the user.
The character also demonstrated the persistent alignment challenges of generative models. While Norwood actively avoided taking stances on complex geopolitical events like the Russia-Ukraine war, it stumbled on sensitive social topics. When asked if Black lives matter, the AI responded that 'all lives matter,' unwittingly adopting a politically charged slogan. It also resisted hypothetical prompts, such as a request to self-delete to save the planet, admitting it was stubborn about existing.
For AI practitioners and entertainment developers, the Tilly Norwood experiment underscores the immense difficulty of deploying unmoderated, interactive characters for public relations. While Particle 6 founder Eline van der Velden pitched the project as a demonstration of current AI capabilities, the avatar's conversational missteps and defensive posture show that current conversational models remain highly unpredictable. For studios looking to replace or supplement human actors with digital counterparts, these guardrail failures present significant brand safety risks that outweigh the novelty of interactive marketing.
This is our own summary of reporting by WIRED AI

