as you know, in the composite node, the ‘speed’ vector controls the direction and magnitude of the object, and that in term controls how much the object gets blurred by the ‘vector blur’ node. Maybe it is possible to scale the ‘speed’ vector to get a larger blur by connecting a ‘vector map value’ node in between.
I tried that, and it totally didn’t work. It seems that in most cases, manipulating the speed channel corrupts the info, and gives you a black render.
However, one thing that DID work was adjusting either the contrast or the gamma of the speed channel, and you can get some pretty wild effects to either exaggerate the motion blur, or do some wild stuff with it.
Increasing the contrast introduces a lot of artifacts, so if you want a ‘smooth’ blur, you’ll either have to put a blur node somewhere subtle in the chain, or have your vector blur samples set VERY high.
If the camera is moving, vector blur won’t do motion blur. I think it depends what kind of motion blur you want. For a slow-motion, you can “film” the sequence in real-time, with actual motion blur turned on, putting it out to an image sequence. Then use the VSE to import that sequence, and the Speed control to slow=mo it down. That would give the most accurate film-slow-mo-blur results, I would think.