Perhaps the most relevant distinction is between "moderation that the users have a choice about" versus "moderation that the users don't have a choice about".

If a site wants to hide all posts/videos that promote some unpopular political belief, or use offensive words, then implementing that censorship as the default user experience is perfectly acceptable, as long as users can choose to opt out of that censorship.

There might be multiple reasons why a given post/video could be censored, and perhaps there is a small burden on sites to tag every single reason rather than mark it for censorship at the first excuse, but I think that a lot of the tagging work could be made the responsibility of the user who uploaded it.

Such a system would hopefully make moot the slightly disingenuous argument that "If sites can't ban political opinions I don't like then they also won't be able to ban spam". Obviously sites would be allowed to put neutral resource limits on users, to prevent DoS attacks.