Wild stab : VSE + Compositing Nodes

Surely, I can’t be the first person to have this totally crazy, bonkers, outlandish idea:

-What if the output of the Video Sequence Editor could be used as an input in the Compositing Nodes Editor.

Anyone has any thoughts on that? Is it do-able?

Not in 2.49b or earlier, not sure about 2.50. However, most if not all of what you can do in the VSE can also be done in the Compositor (and often more successfully), so it’d be a kind of redundant capability imo.

However, you can go the other way – using a Scene strip in the VSE you can place the output from the Compositor in the VSE.

I’ve heard the argument for only using nodes before, but i’m not quite sure I understand how you would do that, especially when you have loads of shots. Are you supposed to calculate offsets everytime you want to move something? Having everything on a horizontal time line seems much simpler.

How would you do that thing with the compositor going into the VSE?

Yes it is, particularly for standard editing practices, cuts, cross fades, etc, but these can still be done in the Compositor – it depends a lot on how you have your workflow planned. I usually use the Compositor to generate effects sequences (either at render time using Render Layers or as a post operation) that I then cut into the movie with the VSE. I rarely need pre-edited footage in the Compositor, but if so, you could simply output image sequences from the VSE and plug them into the noodle.

Maybe the reason you can’t do VSE-> Compositor is to avoid a dependency loop like VSE->Compositor->VSE all in the same file? Putting Compositor output into the VSE seems the more useful of the two, so if a choice had to be made, that would be the logical one imo.

Each Scene can have a Compositor noodle, and its output is via the Output>Composite node – if you enable “Do Compositor” in the Render settings for that Scene, the Composite node is the Scene’s output data. In the VSE, you can add a Scene strip much like you can a movie or sound track – this will reflect the Compositor output of the Scene associated with that Scene strip.

You can always render the output from the VSE to a file, then use that file as an input in the compositor.

In the VSE, just click on Add/Scene. Then a little list box will popup with the names of all the Scenes in the BLEND file. Choose the one you want to add to the VSE.

You will immediately face a problem with this approach, however. The VSE just clicks the render button every time you scrub to a new frame. An immediate bog down will occur if you have anything other than the default cube scene. It makes sense, to me, that there should be a way to get the preview of the scene instead of the full render when scrubbing. Maybe 2.5 will be able to do that…

Cheers for those explanation guys. I can see how things could loop badly with VSE>comp>VSE

In this post, someone brings up the idea of having a different set of nodes assignable to each VSE strip. Now, that would do the job nicely!

For the time being I think I’ll render my VSE strips to a file and then use it in the compositor as you guys suggested.

Thank youse!

Not the first time that ‘someones’ done that, made a few over the years, but it’s all talk ;-). btw your original query has been raised many times, ever since Blender got composite nodes and discussed at length. :slight_smile:

Think it comes down to the VSE being a near application in itself not really interconnected well with the same code as the nodes and perhaps there would be a real slow down in routing back and forth between the two as the coding stands.

There was talk of creating a combined set of code that would handle both VSE and comp nodes processing in the VSE Filter Stack discussions a while ago between developers. Matt Ebb, Xavier Thomas and Peter Schlaile.

Oh well nevermind then…

It’s probably better to have developers focus on the 3D/animation side of things, which is what Blender is all about anyway.
Sadly, when it comes to editing video on Linux, blender’s VSE is the “stablest” tool yet IMO (or at least, the one I’m most familiar with)

Maybe someone could base a new cross-platform NLE project on VSE’s code or maybe that’s the worst idea ever. I just don’t know enough about coding to see the difference :slight_smile: