Automatic speech-to-text subtitling

Why don't one consider making a special program (accessibility, disability) that "patches" to the audio output of any software that contains dictation, such as newsreader, video involving talking, etc, and the software would automatically make an "attempt" to recognise the words and convert them to text, and bring them to the front of the screen, user-configurable to any fonts/colour, etc. I am aware it may not be 100% accurate, but it's worth a try? I have seen software package, that you dictate and it comes up words for word processing, etc, so why not make use of this for converting spoken words into subtitles? The usefulness would be aimed at deaf people, people wanting to listen and read subtitles, etc. Thank you.