If people still can't try out Sora for themselves and determine to what extent examples have been cherrypicked, it would be a bit ridiculous to vote it as better than tools that are actually out and producing impressive results, IMO.

OpenAI just released an awesome new model which makes this question much more interesting