The Use of Text-to-Speech (TTS) Feature on ChatGPT in Developing Students’ Speaking Skills in Terms of Accuracy, Fluency, and Comprehensibility

Abstract

This study aimed to investigate the use of the Text-to-Speech (TTS) feature in ChatGPT in improving the speaking skills of seventh-grade students at SMPN 8 Makassar. Employing a pre-experimental one-group pretest-posttest design, the study involved 36 students selected through purposive sampling. The treatment was conducted over four sessions utilizing TTS-based activities, including pronunciation, shadowing practice, dialogue simulation, and interactive speaking tasks. Students' speaking performance was assessed across three dimensions, namely fluency, accuracy, and comprehensibility, using Heaton's (1988) scoring rubric. The findings revealed measurable improvements in all three aspects: fluency increased from 2.28 to 2.92 (28%), accuracy from 2.14 to 2.61 (22%), and comprehensibility from 2.67 to 3.60 (35%). The proportion of students classified as “very poor” declined significantly from 69.4% in the pre-test to 19.4% in the post-test. In comparison, 8.3% of students reached the “fair” category, and 2.8% achieved the “good” category. The paired sample t-test yielded a value of 10.937, exceeding the t-table value of 2.030, confirming a statistically significant improvement, further supported by a large effect size (Cohen's d = 1.82). Although the N-Gain score of 0.18 indicated a low magnitude of gain relative to the maximum possible improvement, these results suggest that ChatGPT's TTS feature is a useful and accessible tool for enhancing EFL students' speaking skills.