I'm going to look at how some pre-existing implementations work (that I have actually used) - and hopefully that will give you some ideas on how to integrate it with your game engine.
Operation Flashpoint (original release) has a command and control system for entities. This is essentially a list of commands that entities had to perform - the AI system would be responsible for carrying out the commands. Entities could be anything: an enemy, the camera, the player, etc. All the entities had to exist 'before the fact' (but there is nothing stopping you from making a 'create entity' command). Let's take the example of two people meeting, having a conversation, getting into a car and driving off.
- Entity1: ignore player
- Entity2: ignore player
- Camera: Track Entity1
- Entity1: move to Alpha
- Entity2: run to Alpha
- Entity1: start talking, play 'sound1.ogg' when: Entity1 at Alpha when: Entity2 at Alpha
- Entity1: move to and enter Car1 (driver) when: 'sound1.ogg' complete
- Entity2: move to and enter Car1 (passenger) when: 'sound1.ogg' complete
- Entity1: move to Charlie when: Entity1 in Car1 when: Entity2 in Car1
- Camera: restore, Entity1: restore player aggression, Entity2: restore player aggression when: Car1 enters Area1
This system did have some bugs (very rarely cut-scenes that never ended because of border cases like where the AI didn't choose a predictable route) - but resulted in very 'natural' looking sequences because people were not following straight paths etc.
XAML uses a story-board system. This is basically a list of animations that need to play - it is essentially the same as the Operation Flashpoint example except the timings are used instead of 'waiting on things' and entities move along predetermined paths. Again, using the above example:
- Entity1: move from Alpha to Charlie at 0s taking 5s this will be a straight line
- Entity2: move from Bravo to Charlie at 0s taking 5s
- Camera1: move to Delta at 0s taking 1s
- Camera1: move target from (Start) to Charlie at 0s taking 5s
- Entity1: start talking animation at 5s
- Sound: play 'sound1.ogg' at 5s originating from Charlie
- Entity1: move to Echo at 10s taking 5s
- Entity2: move to Echo at 10s taking 5s
- Entity1: get in Car1 (driver) at 15s
- Entity2: get in Car1 (driver) at 15s
- Car1: move to Foxtrot at 16s taking 30s
- Camera: restore at 19s
There are two problems with this (I prefer it less):
- Things could move unnaturally fast or slow
- If you want to change the cut-scene you will need to go and adjust all the timings.
How to do it in XNA
After you have chosen a way to describe cut-scenes you will likely have an some form a file from your editor (possibly an XML file). You would then create a content processor and add your cut-scene to your content project. At runtime you would then call something along the lines of