This paper presents an analysis of quantified characteristics of Disfluency pattern of a leading South Korean film actor Kang-ho Song’s Speech for character design in No.3 (Comedy, 1997), Memories of Murder (Thriller, 2003) and Thirst (Melodrama, 2009). Disfluencies, such as filled pause, repeated words, and repair or repetition utterance, are prevalent in spontaneous spoken language and show remarkably regular trends in a number of dimensions (Elizabeth Shriberg, 1994). The transcriptions of the actor's speech with disfluencies and variables in above films are analyzed, which might affect fluency rates in a corpus. It will be seen that the rates of disfluencies, induced by artificial manipulations, reflect the psychological state of a character. The actor's disfluency rates were relatively increased in Comedy and Thriller than Melodrama. Disfluencies are related with these cinematic factors: movie genre, actor role. Reflecting these features, we then recognize the acoustic features needed for character design.