Very interesting! This looks very promising.
Will you also be able to categorize all the videos? (f.e. food, sports, music) and if that's the case, how will you be able to do it? Do you need tags / titles from the original post or can it be recognized inside the video material itself?
Also, what defines if a video is hot? Is the weight of upvotes/likes for a video equal for all different platforms or is 1 like on youtube less worth then 1 like on vimeo as youtube has a bigger audience?
Good questions.
The DTube and DLive tags are stored in the blockchain, and can be used easily. The original YouTube video tags are collected into our database, but I'm not yet sure whether to use them or the Steem post tags (or both) for searching/filtering.
Currently the hot/trending scoring is based (similarly to Steem), on a combination of the pending payout and the time since the video was created, there is no connection to the YouTube voting at the moment. I'm free to try a different formula because I'm using a separate database, but it needs to be quite efficient to calculate since it must be done frequently (so looking up YouTube votes every few minutes probably won't work), but it's an interesting thought.