While the common way to enjoy two-person comedic dialogue, such as double act, is by watching it, performing Japanese manzai has recently gained attention for its entertainment value and effectiveness in improving dialogue skills. However, performing manzai is challenging, as it requires memorizing scripts and mastering its unique intonation, actions, and timing. In this study, we propose “Manzai Karaoke,” a real-time visual guidance system that assists two-person manzai performances. Similar to karaoke, this system provides script line displays and nonverbal performance cues, enabling inexperienced users to perform manzai. To design an interface that meets the key requirements—allowing users to perform with ease and ensuring audience amusement—we conducted a two-step prototyping and user study. The first step validated the feasibility of real-time comedy performance assistance, revealing that in addition to displaying script lines, guidance on intonation, emotion, actions, and timing is essential. The second step demonstrated that the implemented interface incorporating these elements effectively meets the requirements. This study contributes to the field of entertainment computing by demonstrating the potential of visual guidance and providing design recommendations for assisting novice manzai performances.