For power users needing the highest accuracy, a two-step "Desktop OCR" workflow is standard. The Process: VideoSubFinder:

When subtitles are burned into the video, they become pixels. Your computer doesn’t see “words” — it sees a pattern of light and dark pixels. Extracting text requires an OCR engine to recognize characters, which is prone to errors.

specifically around the area where subtitles appear. This prevents the software from accidentally trying to read background text or logos.