Hi guys,
I am currently setting up a multi-thousand frame scene and I am seriously looking at the benefits, or lack thereof, of motion blur.
I know the in-and-outs of motion blur, so I don’t need help on setting it up, but what I am interested in knowing from people who have rendered up animations with and without its use as to whether it really adds anything to the animation. With a multi-thousand frame scene and looking at a couple of hours per, having the renderer decide to render 5,8,11 or 16 times the number of frames is obviously something I’d rather avoid if it adds nothing or only negligable value to the animation.
I also know that it’s possible to motion blur by effectively half-framing with the sequencer - a question for anyone who has used that method therefore, is it a decent option? (quality wise).
Well, if you have the ability to splice different files together, I’d recommend rendering the whole animation with OSA only on and then re-render the streches you want to be MBlur’d. OSA should be on because it softens edges, just like MBlur does. Otherwise, you’ll get a strange quality jump.
As for the benifits of MBlur, I can’t say there are many for animation purposes. I only use it for movement in stills. If you want, you can fake a blur (and shorten render time) by duplicating, offsetting, and transparenting objects.
Motion blur is definitely helpful for animation, if that’s the look you’re going for. Basically, if you have fast movement and you don’t want the animation to look like stop-motion animation, motion blur will help the movement appear more natural. Now, Blender’s motion blur is a rather rudimentary implementation. It’d be nice to have an image-based motion blur at some point in the future… but I digress. Yes, motion blur is helpful for animation. Also, when rendering with Mblur, you can turn off OSA, because Blender’s motion blur is oversampling… so there’s no real need to do it twice. Depending on the framerate and render size you’re using, you may be able to get away with an Mblur setting of 5. I generally use 8 and rarely go above that (render times get too long for me).
I currently study 3D animation at school and when the subject came upp in one of our animation courses our teacher told ous not to use motionblur when rendering out pictures from the 3D program (we use maya but the same aplys to blender) because the rendering time gets to long and the quality of the motioblur generaly aint to good. But he also said that motionblur basicly always should be added in the postproduction, whitch gives motionblurr that is easier to controll, is faster to rendered and of better quality, this to give an apearance that the viewer is used to see in movies altough it doesn’t exist in real life (like deth of field, it doesnt exist in reallife but we expect to see it in movies and pictures so they look unreal and unnatural without it).
I currently study 3D animation at school and when the subject came upp in one of our animation courses our teacher told ous not to use motionblur when rendering out pictures from the 3D program (we use maya but the same aplys to blender) because the rendering time gets to long and the quality of the motioblur generaly aint to good. But he also said that motionblur basicly always should be added in the postproduction, whitch gives motionblurr that is easier to controll, is faster to rendered and of better quality, this to give an apearance that the viewer is used to see in movies altough it doesn’t exist in real life (like deth of field, it doesnt exist in reallife but we expect to see it in movies and pictures so they look unreal and unnatural without it).
If you separate your ‘moving’ elements - a character lets say, and render them separate from the background, you can save a lot of time. That way only the fast moving elements will get blurred - and you can even choose to blur them with another app.
What puzzles me greatly about Blender’s motion-blur … and I haven’t had the time to try to see how to fix it … is that, when you are rendering an animation, Blender still insists on calculating all of those frames, over and over again.
You see, it’s clear that when Blender is calculating the image for some frame #X, it’s calculating the imagery for frames #X+0, #X+1, #X+2 and so-on (or is it “minus?”) and then merging them together to create the blur. Very well, then, but for the next frame in the animation it stands to reason that Blender should still have that intermediate image-data. It should be able to cabbage it out of some cache somewhere and re-use it, thus having to re-generate only one new frame from scratch.
This would be of no value in generating a single frame, but in the case of motion-blur you’re undoubtedly doing an animation, so it would dramatically reduce the amount of time required.
On another note: I once had to animate a wheel that was rotating continuously. Instead of doing motion-blur, I just rendered the rotating wheel and blurred it in the Sequence Editor. The effect was quite satisfactory, although this is a fairly special case. (Circular motion, and continuous.)
If you are doing “a multi-thousand frame scene,” get to know the Sequence Editor very well. Master the art of compositing. For instance, if your characters are moving against a static background with a non-moving camera, only one frame of background need to be generated for the entire shot. The only stuff that must be regenerated is each motion and the shadows that go with each moving object. I use unix Makefiles for this.
Actually, Blender’s MBlur takes the amount of frames from the BlurFac variable (default of 0.5) and splits them into 5, 8, 11, or 16 pieces. Then it renders those pieces.
Thanks for all of the comments guys. Reading through them has really helped me make up my mind in regards to what is the best trade off.
Unfortunately, the ‘moving element’ in the animation is actually the camera as the animation is being seen through the eyes of the main character - as a result, everything is technically in motion in regards to the motion-blur scenario, and hence why it is such a big problem to me.
I think that I’ll go down the non-rendered motion blur route for this - analysing the frames in post and sequencer to see where I need to add post-process motion blur. With the increased times that motion-blur will require on this, I think that even a relatively slow post-process option will save time the long-run.
