Rule-Based Camerawork Controller for Automatic Comic Generation from Game Log

Key insights

7 takeaways

01
Camerawork rules were derived from an analysis of Lineage 2 webcomics, with one rule per event type covering camera angle, camera position, and zoom position.See it on page 1
02
Shot-type weights come from frequency counts in the analyzed comics, and the controller uses them in a roulette-wheel fashion to pick a shot type for each camera parameter.See it on page 3
03
For fighting events, the camera-angle roulette wheel has four areas in the proportion 65:631:49:0 across Low Angle, Eye Level, High Angle, and Bird Eye shots.See it on page 4
04
The previous controller used only one combination of camera angle, camera position, and zoom position per event type, which produced repeated shot subsequences in the comparison figure.See it on page 7
05
The controller determines an event type by majority voting over the action types occurring in a frame.See it on page 4
06
The authors' evaluation is a visual comparison of generated comic pages, and they state the proposed controller shows richer shot variety.See it on page 7
07
Future work proposes fine-tuning camera parameters based on the entropy of each frame segment, using the Berkeley Segmentation Engine.See it on page 7

Summary

This paper, by Ruck Thawonmas, Ko Oda, and Tomonori Shuda of Ritsumeikan University's Intelligent Computer Entertainment Laboratory, proposes a rule-based camerawork controller for an automatic comic generation system that creates comics from game logs. The camerawork rules were derived from an analysis of Japanese online webcomics about Lineage 2, read independently by two college students using the same analysis manual. The analysis covered five event types: chatting, fighting, moving, approaching, and special. Each event type has weights for camera angle (four shot types), camera position (five), and zoom position (six). The controller selects shot types by roulette-wheel selection using these weights. The authors compare comic frames generated by the new controller with those from their previous heuristic-based controller, using visual comparison on a chatting scene and a fighting scene, and conclude that the new controller offers richer shot variety.

Rule-Based Camerawork Controller for Automatic Comic Generation from Game Log

Rule-Based Camerawork Controller for Automatic Comic Generation from Game Log

Key insights

Summary

Next steps

More on AI

Efficiency of Static Knowledge Bias in Monte-Carlo Tree Search

Player Perceptions of Generative AI in Games: A Steam Review Analysis

Finding Good Work in Speculative Game Development

Stopping Unintended Actions: Exploring EMS-Based Action Inhibition through a Game-Based Rapid Decision Task

Envisioning and Co-Creating Generative AI for Future Creative Workers in Gaming

Orchestrated Reality: From Role-Play to Living, Playable Game Worlds

Sound by Design: Practitioner Perspectives on Factors Moderating the Impact of Audio in Games

NotebookLM: LARP Participant Support Tool Implementation Report

Emotion Dynamics in Social Deception Games: Analysis of Players in Werewolf Games

Preserving Japanese Indie Games at BitSummit Drift 2024

Maximizing Efficiency in Game Development Through Art Styles, AI Integration, and Creative Expression

Level the Level: Balancing Game Levels for Asymmetric Player Archetypes With Reinforcement Learning