[Bf-compositor] Extending some node functions to VSE

[Angus Hollands]

Hi all, this is a very brief concept on workflow, and it might be rather
short sighted.

The VSE is built around layer based, temporally dependent sequences. The
compositor is also aware of time, it simply lacks the layers. To me it
would seem silly to attempt to create a layer node, and provide a sound
node, because this only applies to the VSE, and image compositing doesn't
require some of these features. Moreover. layering is best done as the VSE
does, for now.

I believe that if we wanted to combine the facility of the two in the same
pipeline, it might be useful to define "layer" effect strips which are used
as timing queues for the compositor. When using the VSE if the compositor
is enabled, it will have the ability to composite the entire sequence, as
well as take input from a VSE layer node. This node would serve two
purposes:

1. It acts like a switch, becoming enabled when the layer being rendered in
the VSE.
2. It provides timing information (a relative frame, and the total number
of frames) which may also be of use as the compositor provides switch
statements and a time value already.

This way, the compositor is the final stage of the pipeline, and receives
the output of the VSE using a VSE input node.

This gives the compositor the timing information from the VSE without
attempting to assimilate its behaviour, as they are suited to their own
tasks.

In terms of implementation, this wouldn't require any complex switching
between VSE and compositor, so no nested composites. (that would be ugly!)

It also means that the VSE and the compositor could be used in tandem with
the render pipeline. Currently, it would require the user to shift
animations to the correct frame to match the VSE which seems incorrect. We
should be treating the render output as rendered footage, rather than
immediate rendering, even though it is rendering it on the fly. Therefore,
it would seem sensible that the VSE should either:

1. Ask the Blender render to render specific frames. This seems like a bad
idea because, although I haven't looked at this, I suspect that the time
progression (frame) within Blender is not unique to each editor, and is
global.
2. Ask the Blender render to render the entire scene to disk and then
retrieve specific frames, but do behind the scenes.

Anyway, this was more of a thought exercise than an actual proposal. and
I'm sure I shall be told why this is an awful idea!

Regards,
Angus Hollands
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.blender.org/pipermail/bf-compositor/attachments/20140523/746877ca/attachment.htm