It sounds as if you might have a video file that has text, and you want to recognize the text (OCR, Optical Character Recognition) and convert the text to audio (Text to Speech), and write the result to an audio file. Is that correct?
I would not consider that a project suitable for a beginner. The tools for OCR are not difficult if you have Computer Vision Toolbox, but the easy Text to Speech is by using third-party APIs like IBM or Microsoft to create speech -- and that process usually outputs to the speaker instead of to memory (or file.)
