Abstract
Animations generated by interactive 3D computer graphics applications are typically portrayed either from a particular character's point of view or from a small set of strategically-placed viewpoints. By ignoring camera placement, such applications fail to realize important storytelling capabilities that have been explored by cinematographers for many years. In this paper, we describe several of the principles of cinematography and show how they can be formalized into a declarative language, called the Declarative Camera Control Language (DCCL). We describe the application of DCCL within the context of a simple interactive video game and argue that DCCL represents cinematic knowledge at the same level of abstraction as expert directors by encoding 16 idioms from a film textbook. These idioms produce compelling animations, as demonstrated on the accompanying videotape.
Introduction
In the mid-1990s, while interactive 3D graphics were advancing rapidly, their "cinematography" remained primitive. Most applications relied on rigid first-person points-of-view (POV) or a few static, pre-placed cameras. This approach ignored decades of storytelling expertise developed by human cinematographers. The challenge in automating this process is that film rules are often informal and descriptive (e.g., "don't cross the line of interest"). Translating these heuristics into precise, algorithmic instructions that a computer can execute in dynamic 3D environments is difficult. This paper introduces the Declarative Camera Control Language (DCCL) and the Camera Planning System (CPS) to bridge the gap between high-level artistic intent and low-level geometric constraints.
The DCCL Framework and Camera Planning
The authors' solution involves a two-part system designed to encode cinematic knowledge. DCCL is a declarative language used to define idioms—stereotypical sequences of shots used to capture specific actions. These idioms are constructed from fragments, which are the smallest units of camera motion.
By decomposing complex cinematography into these modular primitives, the system can assemble sophisticated sequences. A critical component of this logic is respecting the "180-degree rule." The system explicitly defines a "line of interest" between actors and ensures the camera does not jump across it, which would otherwise disorient the viewer.
The Camera Planning System (CPS) processes animation data through a three-stage pipeline:
- Sequence Planner: Partitions raw animation data into scenes based on character activity.
- DCCL Compiler: Solves geometric constraints to find all valid camera placements for a scene.
- Heuristic Evaluator: Acts as an "automated director" to rank candidate shots based on visual quality rules, such as transition smoothness and shot duration.
Evaluation and Results
The system was evaluated by hand-encoding 16 complex idioms from Daniel Arijon’s Grammar of the Film Language (1976) and testing them in a prototype 3D game. The results demonstrated that textbook cinematic knowledge could be successfully formalized for computation.
Performance tests on an SGI Onyx workstation showed that the system could plan several hundred frames of cinematography in under a minute; for example, a 733-frame sequence was planned in approximately 47 seconds. Furthermore, while the initial sequence planning was game-specific, the DCCL Compiler and Heuristic Evaluator proved to be domain-independent, making them potentially applicable to any 3D animation environment.
Conclusion and Future Work
While the current system requires custom code for high-level storyboarding, it proves that cinematic expertise can be represented at a high level of abstraction. Future developments could involve more sophisticated heuristics—accounting for factors like lighting or complex multi-actor interactions—to reach feature-film quality. The work lays the foundation for truly intelligent, automated directors in interactive media.
Summary updated