ENHANCING 10TH GRADE STUDENTS’ SPEAKING SKILLS THROUGH AI-BASED VIDEO DUBBING ACTIVITIES: A CLASSROOM ACTION RESEARCH STUDY IN VIETNAM

Abstract

Artificial intelligence (AI) is expanding the possibilities for repeated, individualized oral practice in English as a foreign language (EFL) classroom; however, evidence from Vietnamese secondary schools remains limited. This classroom action research study examined how AI-based video dubbing activities supported the speaking development of 40 Grade 10 students at a high school in Bac Ninh Province, Vietnam. The ten-week intervention followed two cycles of planning, action, observation, and reflection. Students listened to short video clips, used text-to-speech and recording functions as speech models, rehearsed and synchronized dialogue, produced dubbing recordings, and revised their work using teacher and peer feedback. Data came from an analytic speaking pre-test and post-test, classroom observations, students’ products, and a 12-item perception questionnaire. Descriptive results showed that the mean speaking score increased from 11.25 (SD = 2.10) to 15.85 (SD = 1.95) on a 20-point scale. The proportion of students at the good or high proficiency levels rose from 37.5% to 87.5%, and no student remained in the low category. Pronunciation (+1.35) and fluency (+1.25) recorded the largest gains, followed by accuracy and confidence (+1.00 each). Questionnaire means ranged from 4.05 to 4.45, indicating favorable perceptions of improvement, enjoyment, ease of practice, and continued use. The findings suggest that carefully scaffolded AI-based dubbing can provide intensive rehearsal and lower-pressure speaking practice. Because the study involved one intact class, no control group, and descriptive analysis, the outcomes should be interpreted as context-specific evidence rather than proof of causal effectiveness.