The classic Noir narrative structure is a deliberately disorienting mix of two main activities.
One (the frame) features reflective and recollecting voice overs from the protagonist, this quiet, intimate vocal mixed up high over the relatively silent visuals (action, sfx but no important dialogue).
These voice over sequences set up the dialogue and action heavy parts of the noir, mostly events from the past, cast as if they are occurring in the present, in real time.
It all reminds me of the way most music videos work - the sequences of the singer at a mike singing have a boring visual background , if they have one at all. The words, usually of non-repeated 'verses' - reflective and recalling , are everything.
The singer - for example - recalls their lost lover.
In the dialogue and action segments (the so called actors' bits--- even if the singer is now one of those actors), there is frequently either no singing at all (the action occurs against an instrumental solo) or the only singing is the 'voices off' repetitious singing of an already frequently heard (and unsubtle) vocal chorus, more meant to be emotionally felt than intellectually dissected in the foreground.
Here we see scenes of the singer and lover when they were happy and still together.
The song bridge might be the place in the narrative where the nominal cause for the breakup is described in words (voice over - voices off) and seen in the visuals.
In noir or music video it is a very flexible way to produce depth and narrative complexity --- cheaply.
What's there not to love about that ?