It makes the transcode loop easier to read (30% less code)
and the differences with avconv easier to spot.