In the field of human-computer interaction, researchers have explored methods to enhance the turn-taking abilities of conversational agents and robots in interactions with humans. Previous studies have shown that variations in turn-taking patterns (e.g., overlaps and gaps) can influence the perceived conversational atmosphere, and thus, their effect on human perceptions has been examined. However, the combination impact of turn-taking behavior and speech speed on conversational atmosphere remains underexplored. To address this, we developed a conversational simulator featuring virtual conversational characters with simple shapes and meaningless utterances to investigate how the combination of turn-taking and speech speed influences the identification of a conversational atmosphere. The characters followed basic turn-taking rules, and the simulator controlled their turn-taking behaviors. We conducted an experiment in which participants observed scenes generated by the simulator. A two-way repeated-measures ANOVA was performed, examining the effects of turn-taking behaviors (overlap, no-gap-no-overlap, gap) and speech speed (fast, medium, slow). Results indicated that fast speed was perceived as creating a competitive atmosphere, even when turn-taking adhered to no-gap-no-overlap patterns. This finding will contribute to developing conversational agents/robots that have the same capacity to judge conversational atmosphere as humans.