Back in 1983, Dragon's Lair was one of the very first and best-known attempts to mate a cinematic style of narrative with gameplay. Thanks to the fact that the entire game was essentially run by a laserdisc player, this arcade game was able to produce astonishing visuals which would surpass the abilities of "normal" video games for over a decade.
Unfortunately, the limits of its technology meant that gameplay had to be reduced to an even more rudimentary level than other games of its time. A non-interactive video sequence would play, and at its end the player would have to press some input -- either a direction or "attack."
If the wrong input was pressed, the player would see a death animation. If the right input was pressed, the player got to see the next video sequence. Since the game was designed with little consideration for the player's read of the action, the correct input was usually a guess on the player's part; the only way to successful play was by blind trial and error.
Sixteen years later, the Dreamcast game Shenmue was released, featuring a special game mode called the "Quick Time Event" or QTE. During a QTE, buttons would appear on-screen at various intervals.
If you pressed the corresponding button on the controller quickly enough, your character would do the corresponding action. It's structurally identical to Dragon's Lair, except that explicit cues appear on-screen.
What's gained by designing a game this way? Normally, good design principles say that the developer should take special care to present the action so that the player can see what they are doing, what the enemy is doing, as well as accurately apprehend the spatial relationships between all gameplay-relevant elements in the scene.
With this extreme abstraction of player inputs, the developer affords the freedom to display the action as cinematically as possible with impunity, because the action has become so abstract that the player doesn't need to know what exactly their avatar is doing, or what the enemy is doing, or how a change in the rules in this sequence might impact action in another part of the game. In fact, we get to throw out all of the difficult design considerations that we would have to tackle with conventional gameplay.
Something valuable is also sacrificed as a side effect of this ultimate abstraction of control. Since we threw out all of the difficult gameplay design, we necessarily decoupled the gameplay from the action so much that the player in fact does not need to pay attention to anything other than the on-screen cues of what button to press. If we displayed just a black background, in place of a cinematic sequence, the nature of the interaction would remain unchanged.
God of War, released six years after Shenmue, employs a virtually identical system at certain gameplay moments. While the cinematic quality of these sequences are breathtaking, and certainly are leaps and bounds above what either of our previous examples could accomplish, on the gameplay level we see the exact same kind of decoupling and rudimentary interaction.
Inis' Gitaroo Man, released in 2002, offered highly cinematic narrative as a background, operating in concert with a sophisticated rhythm-action gameplay system in the foreground. This is an interesting example because even though the gameplay feels very connected to the narrative, the two were totally decoupled, in the sense that the narrative didn't contain any information that would help you play the game effectively -- like the previous example, gameplay interaction was unaffected by anything going on in the background.
However, the background was affected by the gameplay, and this resulted in a very nice illusion that made two independent layers feel like one. The gameplay patterns were scripted to match the music, and the background narrative was also scripted to coincide with the music track. The player character and enemies would also pose and act according to your inputs, although you didn't need to notice or observe them at all in order to play the game. The developer's DS game, Elite Beat Agents, uses many of the same techniques.
What we can learn from these examples, at least with this technique, is that the developer gains the freedom to do anything they want with their game camera, and indeed the background action, at the cost of severely confining the quality of gameplay to "press this now" type of interaction.
In other words, when we allow unrestricted cinematic narrative expression during gameplay, the gameplay itself must be either severely limited or decoupled from the narrative.