I'm ok with this, but the user base is so little, who's willing to pay within those? Some might just change front end...

True, but we don't have to cover the cost of a 10+ developer team :D

As for image storage, what if instead of stare the image somewhere, you encode the image as a string in the post or comment and then decode when showing it on frontend?

There is a limit to the size of a post (less than 64k bytes). You cannot really encode good images within the post itself.
And even if possible I think the witnesses will start complaining (correctly) because we are basically offloading the cost from the frontend to the chain block processing.