@mppic: still a little bit confused. The speed controller can be configured using IPOs to do only blends or strobes or both… ?

Maybe I wasn't too clear about what I'd like to achieve. I have two strips that I want to have a similar look: First I stretch an image sequence via speed control, frame blending turned on - everything fine so far. The second strip is a single image which I zoom into via transform, which (correctly) gives a smooth animation. To make this strip look like the previous (with blending), I'd have to 'drop' some of the frames transform calculated and replace them with blends, which isn't possible right now (correct me if I'm wrong).

I realize now that integrating frame blending in the transorm tool is more complicated than I originally thought because you'd also need some sort of 'strobing' or 'step size' that tells transform *not* to create all inbetween frames.

A more general approach would be to turn strobe and frame blending into independent effect strips. You could then take transform output, drop frames as needed with strobe and after that blend the remaining frames.

Quite flexible, but even more complicated to implement - and maybe there isn't any other 'need' or use for this except my case, so I guess we might as well forget about my ideas ;-)

@Peter Schlaile:

"I read the Gnonlin sources. The seeking is handed over to the input plugin by passing up a seek message through the tree. The input plugin then performs the seek operation. This is rather naivly implemented and will certainly not work with mpeg files. Looking at my latest version of the seek code maybe it is possible to work around this."

Ok that does seem a little silly, and I don't even program this type of stuff. I'm not familiar with this type of thing, but I would think that it would be better to have something like a ring buffer set up that makes calls to the input plugin to get frames to fill the buffer with, and then do the seeking on the frames within the buffer. (I'm sure that's not how it's actually done, but even I know that seeking mpeg files is a pain)

I'm sure they're always open to suggestions and tips on how to better implement the seeking, especially since Gnonlin is targeted towards audio video editing programs.

"That sounds interesting. Is there example code out there, that shows that?"

Probably, but I would imagine the folks on #gstreamer at freenode would know a lot better than I do where to point you for that type of thing.

So, perhaps a workflow could be done more similar to filming a live-action movie. When you're still in the animatic stage with very basic models and animation, you could create longer takes and even multiple takes of the same action exploring different ideas. At this point, you can depend on the asset management and the strength of the editing tools to create a relatively solid edit of the whole movie. Then the pro voice acting can be done. Now you can go back and flesh the portions that were actually used.

So, in this case the power of the editing suite would be more important because it would be tightly integrated into the process of creation instead of just assembling finished pieces of animation at the very end. In this kind of situation, interesting integration options could possibly be done. Like if you select a clip in the assets or timeline and do a special option, it could open up the scene in blender, marking the ranges of frames that are actually being used in the final edit.

So, I think a powerful editing section (including assets) is actually more important for a project like Elephant's Dream than something like a live-action film that happens to have one 3d scenen in it. In the latter case, it may very well be better to use a dedicated video app which is more likely to have video-related features like de-interlacing, capturing directly from a DV source, etc. (though the lack of a good open source editing program on windows is still pretty glaring). But for the former case, it could potentially save a lot of time by having a version of the final edit throughout the whole project to make sure that a final render only happens for what is absolutely needed.