I never used bink, so I can't tell you any specifics on that part, but if you want multiple objects to be played as a video you will need to render the video to a texture and make the video play the relevant parts according to the interaction.

However, I think you mean to have a video you can interact with. If that is the case, you probably need to create a video for every interaction (not necessarily multiple files, but different scenes) and play the relevant parts of the video accordingly. I don't know if this is the way older games have implemented this, but I am sure someone here knows.

As for the no loading part, I guess it depends a bit. The more content you need to show, the more memory you need and you only have so much of it to be in memory at a time. You can of course be smart and anticipate what parts you need and pre-load it and unload unneeded parts etc or stream in the needed parts.