For Greece and Rome this is the simplest explanation: it is a long narrative written in hexameters (or a comparable vernacular measure) which concentrates either on the fortunes of a great hero or perhaps a great civilization and the interactions of this hero and his civilization with the gods (Merchant 1971: vii). A little more thought suggests a contrast between the type of epic which was passed from generation to generation by word of mouth ('oral' or 'primary' epic such as Homer's Odyssey) and the epic which was composed with a pen ('secondary' or 'written' or 'literate' epic such as Virgil's Aeneid). But such distinctions are very crude. Lucan (AD 39-65) wrote an epic poem entitled The Civil War. It is long, it is narrative, and it is in hexameters, but it has no hero, no gods, and its regard for 'civilization' is scant. Hesiod (c. 700 BC) may have written an epic poem entitled The Shield of Herades. It certainly has a hero, gods, and a narrative, and it is, in its way, about civilization. Yet it is extremely short.
Are there surviving ancient discussions of epic? Koster (1970; compare Thraede 1962) attempts to reconstruct some. The most influential was by Aristotle. Aristotle, in Poetics chapters 23 and 24, specifies some of the characteristics which an epic ought to show. Here are some (I am following Halliwell 1986:257f.; compare Heath 1989:38-55): it must have a plot structure which is 'dramatically' put together; the plot should present a single action 'with beginning, middle and end'; epic should have a unity that is not merely temporal or sequential, nor produced simply by concentrating on a single hero (compare Heath 1989:38); an epic plot ought to be 'compact enough to be grasped as a whole unit'; an epic, like tragedy, should contain