In November 2018, Studio Tender Claws launched Tendar, the first long-form, augmented reality game to merge AR technology with human sentiment analysis. Gameplay centers around an artificially-intelligent pet fish that responds to the player’s actions, emotions, and physical surroundings via a combination of generative and hand-scripted dialog. The fish recognizes over 100 user emotions and 200 physical objects as it navigates eight distinct developmental stages. As a small independent game studio, our challenge was to generate and trigger engaging dialog for the vast combinatoric writing surface that the game presented, all while dynamically adjusting tone, affect, and content according to player actions and the fish’s emotional state.
To address this challenge, we adopted Dialogic, an open-source scripting language and toolkit for interactive, generative dialog. Dialogic, authored by panellist Daniel Howe, integrates generative and scripted content, allowing NPCs to respond organically to non-sequential input from human users. Because the system is open-source and under active development, we were able to adapt it to our needs as they emerged throughout the game’s development. The system proved both versatile enough to be used by our mixed-background writing team, and performant enough for runtime execution in our Unity/Android environment.
This panel brings together Samantha Gorman, co-founder of Tender Claws and lead writer for the project; Ian Hatcher, a member of the core writing team; and Daniel Howe, the creator of Dialogic. Together we will discuss how iterative design and close collaboration between the various teams helped us achieve project goals for both Tendar and Dialogic. We will also present the strategies, processes, and tools we found to be most useful in addressing the vast combinatoric space that the project presented.