Watching subtitled video requires attention distribution between moving images, speech, and written text, yet little is known about the earliest stages of subtitle processing. This study examines how viewer (L1 vs L2) and subtitle length influence subtitle skipping, two latency measures (time to first fixation and processing latency, the latter indexing the initiation of reading after initial corrective saccades), and early saccadic eye movements at subtitle onset (refixations, regressions, standard and corrective forward saccades, and vertical corrective saccades). In an eye-tracking experiment, ninety-five hearing participants (37 L1-English and 58 L2-English) watched English-language clips with same-language subtitles. Subtitles shorter than 15 characters were more likely to be skipped in both language groups, raising concerns about their readability. Both latency measures increased with subtitle length. Additionally, L2 viewers moved their gaze to longer subtitles more slowly than L1 viewers. L2 viewers also produced fewer corrective forward saccades and more refixations than L1 viewers. Finally, longer subtitles reduced the likelihood of refixations and vertical corrective saccades in both groups. These findings demonstrate that the earliest stages of subtitle processing are constrained by both subtitle length and viewer language, with implications for subtitling guidelines, accessibility, and language-learning applications.