Cool idea, my first impression is that there should be some kind of mini-arcs or long phrases that make the content more digestible. What I mean is that now the speaker goes on and on, and the subtitles are abruptly changed to another set of words. As a learner, I would like to have better sense when the words will disappear. I noticed it's arbitrary Youtube content so it's hard.. A simple testable solution for arbitrary content: option to _automatically_ pause/slow down the video just a for second before changing to the next set of text. Another simple and stupid idea would be to show previous words below the the current context.

In the long run usage of a particular user, maybe the app should not highlight words that it knows you have encountered in previous videos many times or if the system knows you have mastered that word -- in case the system is testing user, did not notice if it does.

Btw. Not sure what is the difference between green and black colored words. (Don't know Spanish so hard to guess.)

Also there's some typos at the landing page, at least these: "Activley", "wont be"

Seems like a great concept, good luck :) (I don't find Duolingo that effective either.)