Well, let's step back a little.
What is a person doing when they interact with the image based control?
Making a "selection". (what does that mean? we'll get to that later)
What is a person doing when they interact with that button?
Personally, I would say, they are confirming their selection they previously made, thus one potential wording would be something like "Confirm Selection". The primary alternative, to me, is that they are now "Done" with making their selection.
I try to avoid using "Done" as a single word by itself in multi-stage tasks, as it sometimes becomes contextually unanchored: Am I, as someone using the application, "Done" altogether, or just "Done" with this stage? "Done with selection" or "Done Selecting" is one way to word that with a firmer specific context.
"Save Selection" also works in a modal context, but it also carries some implicit nuance where someone may then expect to be able to later edit their selection.
If there is only a single video being selected and there is a way to return to this selection task screen (back navigation of some kind, or editing this selection later), then I would definitely agree with @N.H.'s answer about removing this button altogether. Interacting with the video selection control is plenty to indicate intent.
What about this screen's prompt itself?
But before I let it go at just that: I really feel like you may wish to back up even further and consider the language of your prompt. "Select a video" is very generic, and presumably worded from the perspective of the system (e.g. what do I, as the developer, need "my users" to do for me at this point), rather than what the person using it hopes to achieve/is engaged in doing. Are they here to choose a video to watch? Are they here to choose a video to edit. Is this part of a survey and they're indicating which video they preferred?
What is the person doing that brought them to this screen? What is their intent? Why are they here? What are they indicating with each control interaction (versus what you are having the system do as a reaction)? Flip the perspective to that and a lot of the language becomes much easier to discover.
It's hard to give better advice about possible alternatives for your task completion control when the task itself is so generically labelled. I'm inclined to think that if your prompt were worded in a way that was more anchored in the person engaging in the task's perspective, the completion control would be easier to word accordingly and sound more contextually specific rather than so generic, which seems to be one root of your question here.
But is "Select" okay as completion button text?
Generally speaking, you certainly "can" name your submit button "Select". But you can also just as well name it "Ok", at that point. I would even say that if you are ending up in a situation with generic terms for signaling task completion, you may as well go with whatever is the most standard for the sake of global consistency, rather than trying to pick something else for possibly the wrong reasons. "Continue" is also common in many multi-stage task interfaces like wizards, surveys, etc.
At the very least, though, "Select" does match the language used for the screen's prompt, so that's internally consistent. At the same time, if that seems generic, maybe that means your focus is in the wrong place, given where the button's text is actually deriving from (the task prompt).
To me, arriving at this question points to needing to back up a step in how you're approaching your language use and crafting an interface that's intended to focus on the perspective of the person using it rather than the data/input needs of the system from a development perspective. Rewording the task prompt from the appropriate perspective would likely inform a better corresponding wording for the task completion control.