The visual channel is far more important for understanding speech than most people with normal hearing realize. Take a video recording of someone speaking. Add noise to the audio until the threshold of intelligibility. Around the point where many words are lost and it's hard to follow, but you can still catch snippets and hear at least parts of most words. Listen to the audio for a bit. Then switch on the video too. Many people will suddenly understand the speaker effortlessly. It's quite jarring if you're particularly susceptible! Similarly, visual information will sometimes take priority over what is heard when decoding speech if you set it up so they conflict: https://en.wikipedia.org/wiki/McGurk_effect