Training dataset from mp4 videos + Json objects?

Hey, I’m new to tensorflow although I’ve played around with image classification using python before,

I was wondering if it’s possible to input mp4 videos which also have related json objects which show hashtags, titles, usernames and likes etc for that video to train tensorflow with? I have a lot of these videos saved already (5,000+) along with their corresponding json and I wanted to try to train a neural net to see if it would pick similar videos that I would when inputting an mp4 and it’s json. Greatly appreciate any answers on whether this is possible and would work and pointing me in the right direction!