Although measures of lexical complexity are well established for printed texts, there is currently no equivalent work for videos. This study, therefore, aims to investigate whether existing lexical complexity measures can be extended to predict second language (L2) learners' judgment of video difficulty. Using a corpus of 320 instructional videos, regression models were developed for explaining and predicting difficulty using indices of lexical sophistication, density, and diversity. Results of the study confirm key dimensions of lexical complexity in estimates of video difficulty. In particular, lexical frequency indices accounted for the largest variance in the assessment of video difficulty (R2 = 0.45). We conclude with implications for CALL and suggest areas of further research.