Delayed CC is definitely an interesting idea!
You can set a video player to delay subs - it's absolutely awful. My spouse prefers having the subs in almost all cases - helps mumbling (talking through their asses as it being called) profoundly.
And from someone who is used to watch with subs for his entire life, you get used to it and won't notice them anymore at some point in time.
I had no trouble with this in the game, but in the show it was very distracting. Probably because in the game I could change where I looked to focus on the text of I wanted to.
That’s what I do with VLC. Just press J (or K I never remember) a few times. Around 1s it’s exactly what you describe. You can then occasionally glance after the delivery if you missed a word.
The actors are jumping around erratically and we track them fine. Even if the text always appears in the same place, it still comes at unpredictable times (temporally erratic). And even normal cc moves around within an area anyway. We track all of those randomizations just fine, without thought or effort.
I don't know until someone tries it, but I can imagine it becoming perfectly natural and someone used to that would find the idea of having to look away to the text maddening. It wouldn't (well, might not) be perceived as jumping around but the opposite, placed already where you are looking.
Even when the text does have to jump from where you are looking, because a new speaker spoke, even that might be better than today, because it draws your attention to the new speaker or i terruption. I often actually get confused by normal cc because all speakers are rendered identically and it's not always clear who said what, and sometimes not even clear that a given string of words is from 3 different speakers. By the time you puzzle it out the whole scene is gone and you're now about to miss some new dialog that's about to disappear before you've started interpreting it.
It would hinge a lot on what they said about getting it right. Surely you could make this idea terrible. But surely it's at least possible to imagine a version that isn't terrible. Different fonts, sizes, colors, placement, borders, etc could all combine to make something far more natural to ingest that what we have now.
Like a better version of comic book speech bubbles. No not literally exactly like them. But some aspects of them, such as how they don't all look identical. Thoughts look different than speech. Yelling looks different than speaking. Emotion and intent are conveyed as well, etc.
> Maybe something like using AI/ML to dynamically place the CC closest to the person speaking so I don't need to direct my eyes at the bottom of the screen and can have both the text and the actor's face in my field of vision?
I think this would be even worse, at least for me. Now not only are there distracting words on the screen, but they're also jumping around erratically!
I wish there was an option to delay closed captions so that they appear after the line is spoken, rather than before. That way, I'd be hearing the line before reading it, rather than the other way around, so it doesn't ruin the delivery. It would also make it easier to ignore captions when I don't need them, because I could just glance to the bottom of the screen after I miss a word rather than having to preemptively read everything just in case.