Text to speech models like Whisper are getting good enough that screen readers would stand a chance. The video itself is still more difficult to caption, but the things that would be shared in video are probably being shared with screenshots right now, so it would not be worse.