Various external media files are opened. You play, cue and examine the media. Tagging, labeling and adding notes. Marking of interesting points and ranges. Possibly breaking down into clips, or at least extract some ranges as clips. Draft arranging the clips, applying some effects to check the result and thus to find out about the viability of the footage. Playback of several media at the same time (several videos, but also video and music). Grouping of assets (media, clips, effects, markers) into folders.

You build up a simple linear cut sequence. Either by

Sound is either used immediately as-is (the soundtrack attached to the media), or there is a similarly simple, linear music bed. Some people prefer to switch sound off entirely for this kind of work. In any case, the link is either automatic, or rather vague and soft (as music being vaguely correlated)

Without the intention to rework it from scratch, an already existing simple assembly is augmented, beautified and polished, maybe to conform with professional standards. This includes the “rescue” of a somewhat questionable assembly by repairing localized technical problems, but also shortening and re-arranging, and in extreme cases even changing the narrative structure. A distinctive property of this usage scenario is that work happens rather in the context of tasks (passes) — not so much isolated operations:

Here I define compositional work as a situation where you deal with multiple more or less independent sequences going on in parallel, similar to a musical score. Frequently, we encounter compositional parts embedded in a otherwise linear work, and often those parts evolve when Scenario (3) is driven to the extreme.

The common distinctive property of all those situations is: objects are embedded into a primary context and have to obey the rules of this context, and at the same time have a close correlation to other objects which are embedded in a completely different (“orthogonal”) context. (To give a catchy example: assume, a CG monster has to be integrated. Besides the masked monster object, you have several colouring and blurring layers at completely different levels in the layering order, and at the same time you have correlated sound objects, which need to be integrated into the general sound-scape. And now your primary job is to get the movement and timings of the monster right in relation to the primary timing grid established by the existing edit)

The working style and thus the tool support necessary for compositional work is completely different to Scenario (3). After an initial build-up (which often is very systematic), the working profile can be characterized by tweaks to various parameters to be done in-sync at widely separated sites within the session, together with repeated cycles of “do it”, “assess the result”, “undo all and do some small detail differently”. Typically there is the need for much navigation (contrast this to Scenario (3) where you work in tasks or passes)

The degree of integrating sound work is worth discussing. Often, due to limitations in existing tools, sound work is done in separate applications to a large extent. Which in turn forces the whole production into a sequential organisation scheme. First the edit has to be roughly final, and then the sound people can step in. (Of course this is an simplification). To list the common operations:

While clearly some of those tasks are always better done within a dedicated application, the ability to carry out this work partially within the main session and even while the basic edit is still in flux — may open new artistic possibilities.

At first sight, the operations and the work to be done in large projects is the same as in small ones. But large projects tend to create sort of an additional “layer” on top of the usage scenarios described thus far, which will “kick in” at various places.

Several people work on a project.

The application is started “headless” (without GUI) and controlled via an API. Either an existing session is loaded, or a new session is created and populated. Then, some operations have to be done in a systematic manner, requiring a way to address parts of the session both unambiguously and in a way easy to access and control from a programming environment (you can’t just “see” the right clip, it needs to be tagged). Finally, there might be an export or render step. A variation of this scenario is the automatic extraction of some informations from an existing project.

Well, after having considered, compiled and written such an concept, altogether avoiding a big picture view of the application is not longer an option. To the other extreme, we neither have the resources, nor the circumstances for doing a rigid and formal analysis. Finally, the XP approach really sounds promising, and it should be clear that it is in no way ruled out. Nothing hinders us to have a detailed vision, but then to implement small usage stories which fit into this vision.

Besides, another consideration. The above compilation builds upon the notion, that there is a common denominator of film making craft, a core editing art, which has been shaped in the first 100 years of cinema, and which won’t go away within the next generation, even if the technological and practical circumstances of production change quite dramatically.