The current study investigated the effects of synchronous computer-mediated communication (SCMC) mode (video vs. audio), interlocutor familiarity, second language (L2) proficiency, and task repetition on learners' task engagement as well as the quality of written task outcome. Specifically, learners' cognitive, behavioral, emotional, and social engagement during the performance of collaborative blog post writing tasks in video chats and audio chats were examined. Twenty pairs of Korean learners of English completed the tasks in video chats and audio chats using Google Docs and Zoom over two days. Cognitive engagement was operationalized as collaborative writing episodes (CWEs) on grammar, vocabulary, mechanics, and content. Behavioral engagement was measured using oral word counts, oral turn counts, and written word counts. Surveys measured social and emotional engagement, and interviews provided in-depth insights into learners' engagement. LME models revealed a significant interaction between SCMC mode and a pair's L2 proficiency for cognitive engagement. Pairs with low L2 proficiency produced more CWEs on grammar during video chats, and interlocutor familiarity positively influenced social engagement. Furthermore, task repetition led to a higher quality of written task outcome. The findings offer insights into understanding the relationship between SCMC mode and learner engagement during task performance.