Fighting Game Commentator with Pitch and Loudness Adjustment Utilizing Highlight Cues

Watching video game live-streaming via platforms such as Twitch and YouTube has snowballed in popularity for a decade, In traditional sports, some research investigated commentaries and it has become a new kind of entertainment [1] with a considerable by human commentators from the perspective of their market value [2]. The game commentary can keep game phonetic variation [5]. Commercial games in recent decades, live-streaming audiences entertained and informed [3]. However such as NBA 2K series and FIFA series, commentaries to employ human commentator is costly, and therefore, the by human commentators were pre-recorded then replayed demand for non-human or AI commentators has been surfaced during the gameplay. Therefore, the demand for building and increasingly gained interest from researchers [4]. As game intelligent live commentary generating systems for video commentary is a kind of expressive speech, synthesizing game game live-streaming, expected to bring higher productivity commentary requires not only synthesizing realistic speech at lower costs than human commentators, has surfaced and using text input from game scenes but also adjusting the gained much interest by researchers [4], [8]-[10]. The recent phonetic variance that expresses the emotional information advancement on neural models accelerated the development based on the context [6].