On my admittely low-end (- WinXP Pro - 1.2g / 256meg - GeForce 4 MX 420 / 64mb) p.c. the sequence editor image preview brings my machine to it’s knees. Just clicking anywhere in the sequence editor strip view brings up
an hourglass for at leat 3-5 seconds.
This is the same for 2.41 released and the latest CVS. and CVS/FFMpeg.
I wonder if this is particular to my machine (video card) ?
Also when playing the animation (alt-a_, the test animation
just shows the white-plane moving in 3 steps - bottom left corner
middle of the screen, and top left corner.
Interestingly, when I start the playback using the timeline, the
playback is still extremely slow, but the playback advances by
two frames at a time at approx 1-2 fps.
I even reduced the SizeX, Sizey down to 40x30, and the animation still won’t play smoothly in the preview window.
I tried it on my Winblows XP Home 2Ghz 512MB RAM with an ATI RADEON x600 126MB video card.
The alt-A gets me a jumpy white square (with fringes?). The play button creates a jumpy first run then smooth repeats. Clicking the sequence editor goes frame to frame without a pause. Hope that helps.
Thanks for the feeback. Your specs are roughly double mine, and still there seems to be room for improvment / optimisation of this. That is about as basic a “scene” as you can get
When you have a moment, try bumping your SizeX, SizeY up to whatever your vid card can handle, or until you can’t get it play smoothly.
I also wonder, with this admittedly un-scientific test, if a more complex scene / more strips, translates to a directly linear increase in drawing feedback time?
If any of the programmers read this thread, I’m curious what the basic core difference is in how this preview image window is drawn vs the alt-a animation in a 3d viewport?
I think the reason it’s so slow is that when you put a scene on the timeline, the sequence editor mathematically has to render each frame as you advance the timeline forward/backwards.
This isn’t the same as working with an AVI whereby the image data already exists within an AVI container, and all the sequence editor has to do is locate the next/previous frame. :rolleyes:
I could be wrong, but that’s how I’ve always interpreted how the sequence editor deals with scenes and actual movie clips.
As I understand it, when playing a movie file the computer only deals with the pixels that change from frame to frame. In this simple test you have a large white square move across a blue background. The sqaure is so large, it takes up so much of the frame, that playing the resulting movie means having to process a large percentage of the pixels for every frame. Add to that the stark contrast and the eye can see every little jump or jagged edge.
In other words: this may not be the best test of your computer.
With that said, I got the speedy little square up to 640x480 before it became what I would call unacceptable.
Well, I’m not testing it because my computer is too fast, and I could probably get a normal result. It could be that your graphics card doesn’t support opengl enough.
If you use a more complex object/scene for this test, I think you’ll find that the sequence editor gets even slower still, because it IS rendering the scene.
In order to show you one frame of the scene, it has to render it because there is nothing for it to show you until it renders what it is you’re animating.
Like I said, this is different to using an AVI or sequence of image stills because the images already exist and were already rendered into files.
If you render that ‘square’ file into a proper avi or sequence of images and then put THAT onto the sequence timeline, I think you’ll find it plays back in realtime.
Oh man. Start doing 6 Scene renders with transitions and alpha masks and crossfades, HDR light probes on each Scene, etc…
Moving just one frame can bring a box “to it’s knees” for over an hour on any machine. Because it’s rendering that frame for each, at the renderer’s current settings for each scene… then compositing.
I didn’t realize the preview was being fully rendered .
I setup a slightly more elaborate scene with ray traciing / mirrors / shadows, the image preview is pretty much the same time as a “regular” (F12) render.
What I think would be a terrific feature to implement would be this ;):
When you are in the sequence editor, you present the user with two choices when dealing with SCENE strips: Previs or Production.
When in previs mode, the sequencer rather than rendering an entire frame, simply does a quick snapshot of the camera’s viewport. Much like making a preview video with CTRL-F12.
Since the sequencer is basically accessing what is already being computed/generated by the GFX card, that should be a lot quicker than frame rendering. And useful for getting edit timings correct.
Once satisfied the edit flows correctly, the user could then toggle the sequencer option to production mode, whereby the sequencer is instructed to render a full production quality timeline.
)