You can split test any element of the page - headline, video, body copy, call to action, etc. Just make sure you only change one element at a time for each test - otherwise you won't know what change caused the difference in results. I would personally start with the headline - this is basically the ad for the page and based on the strength of the headline, the reader either will continue reading or hit the back-button. It is possible to see 10-20 times different results it you pick a killer headline.
If you are looking to test just headlines and such, run an mutli-variant test. That way you can test 50+ combinations simultaneously. Split testing is more suited to larger differences like page design. We test everything, ranging from simple headings, images, pop ups, layout/design, calls to action, etc.