The aim of this study was to evaluate the effects of various realistic levels of talking-head on students' emotions in pronunciation learning. Four talking-head characters with varying levels of realism were developed and tested: a nonrealistic three-dimensional character, a realistic three-dimensional character, a two-dimensional character, and an actual human character. The student's emotional level was measured with the learning component of Achievement Emotions Questionnaire following their exploration of the instructional material on a self-paced learning method. The research method was a quasi-experimental design, and the data were analyzed using ANCOVA test. The sample consisted of 150 Semester 1 students from four community colleges. The findings revealed significant differences (p < 0.05) in emotion test outcomes among groups who received different levels of realism for the talking-head in pronunciation learning app. In conclusion, this study recommends a nonrealistic three-dimensional, two-dimensional, or actual human character as the most suitable design for the talking-head in instructional materials.