Mocha 5 more likely to stop tracking?


I’ve been using Mocha 3 pro intermittently for a while and recently upgraded to the Mocha 5 Pro plugin (in order to do away with importing/exporting clips from premiere and Resolve).

In my initial testing I’m finding that this version is far more likely to stop tracking a shape than V3. I almost never had Mocha stop - the shape might stray off on a difficult track but it would not stop. In V5 it seems to stop quite easily and then I have to go about manual tracking or making a holdout matte for an object which partially occludes my tracked object/shape.

Am I missing something here? Is there any setting which makes it stop rather than drift off?



The only difference between v3 and v5 planar tracking would be the GPU. Do you have GPU enabled in your prefs?

Yes I do. Also one of the main reasons I upgraded. I have a K5200 which is a supported GFX card.

Hi there,

It sounds like you’re running out of memory to process the cached clips, which we are very close to having a fix for. Please stand by, and we should have a solution in the next free point release.


Which version number of mocha V5 are you currently running Mike?

I’m running 5.2.1.

Working with HD footage and I have 24GB RAM - it didn’t seem like I was running out of memory. When I get a moment I’ll try tracking the same shot in both V3 and V5 (with and without GPU on) and see what happens.

Do you get the same issue when tracking in a lower proxy?
Are you on Windows or OS X?

I spoke to Jon at Imagineer last night and it seems there are a couple of things going on:

In my specific example I’m trying to remove dirty marks on the lens so when I track a background I am unlinking the spline from the track so that the spline stays in the same place on screen. Apparently the algorithm for the CPU and GPU tracking is slightly different so it is possible there might be some discrepancy between the two. Also, my ‘Min % pixels used’ defaults to 90% (for some unknown reason) and lowering this value does seem to help. Additionally, when you change this parameter in the middle of a track it creates a keyframe which you cannot see on the timeline, you have to dig down in the dope sheet to see it. I think this has been causing additional difficulties in my troubleshooting because I was trying to track using different values for this parameter, unaware that I was actually keyframing it all over the place.

Regardless, my results do seem inconsistent. Several times whilst I was on the phone with Jon, I tracked the shot using GPU and it would stop at a particular place. I would then close the interface, open it again, re-track and it wouldn’t stop. Weird.

I have left my example shot with Jon and hopefully Mary and the engineering team can replicate my issue. Annoyingly it’s inconsistent so tricky to pinpoint.