See, that smell just sort of travels everywhere.
Actually, you're half right about this. What actually happens here I believe is that he mixes up
two different thoughts, and as a result neither one comes out properly. You might write these two thoughts as each consisting of two pairs of sentences:
1)
See that smell? [It] Just sort of travels everywhere.
2)
See? That smell just sort of travels everywhere.
He pronounces the first three words (
see that smell) in very quick succession and then makes a long pause before continuing.
I strongly suspect that he started out trying to say the first part of thought 1 and then, during the pause, probably realising that the idea of 'seeing' a smell doesn't make a lot of sense, attempted to repair the sentence by predicating what was previously the
object of the sentence (
that smell) to become the
subject of a new sentence, i.e., thought 2.
If you pause the video at 4:55, you can actually see this reparation happening. He releases eye contact from the girl he's talking to, and for a split second his eyes defocus and go down and to the right—clearly entering 'search mode' as I would call it. He's monitoring his speech here. It's only as he raises his eyes to lock eye contact again that he proceeds with the newly repaired sentence.
GS—if you're interested, you could try
this different video: go to 5:26.