Tracker auto update reference button

Just curious if anyone has issues with the single tracker as I am copying it between shots then try and re track in a new shot, but with auto update on it tends to jump a few pixels on the first frame and just curious if anyone has a rule of thumb when to turn auto update reference on or off, if I switch it off it tends to not jump anymore, the shot doesn’t have a big move and not sure why the jump occurs but switching this off tends to help.

Auto update means that it uses each previous frame as a reference to compare to. With it off, it uses the original frame that the ref box was placed on as a comparison. The fact that you are reusing an existing track and modifying probably confuses it a bit. It probably doesn’t jump if you are starting each track from scratch, although I have seen it do unusual things on the first 2 frames on occasion. I usually ignore it since I almost always work with handles. On occasion I will start my track a few frames in and run it both forwards and backwards which helps eliminates the occasional weirdness.

yea probably the case but I also noticed that it just doesn’t track as well with more static shots and tends to drift when turned off it drifts less, which is odd since you think if the algorithm works it should be more accurate not less but that’s not what I am seeing on the shots I am working on.

For shots that don’t move much, have auto update turned off. For shots that do move a lot, turn it in. That is always my rule and it works pretty well.

1 Like

I have seen this too. It’s a puzzle. I’ve never stopped long enough to figure out why. Thankfully it’s all hidden in the handles.

This is also my belief. Like it somehow has a memory of the earlier detail and it keeps trying to find that.

The tracker is mysterious. Some times it’ll hang onto the mere idea of a point, pulling a track so perfect you don’t know what to do with the rest of your day, other times it will leap off the most clear piece of detail and settle energetically in a detail free gradient.


makes me think the auto update button needs a review under the hood, now if that will happen I am sure they all have bigger fish to fry but its nice to know I am not the only one that this happens to and good to know some when to use it and when not to use it, thank you kind folks in the Logik community

Hi All,

I know I am late to the party but one thing that has not been mentioned here which might explain the shift on a static shot is the grain. If there was zero grain than the tracker will stay locked to the pattern which no uncertainty. As the pixels display minor value shifts because of the grain, this could have an influence on the tracker. This could be alleviated by adjusting the tolerance of the tracker to account for this situation.

It’s an interesting thought about only using the auto-update reference for moving shots but the original purpose of it being on or off has no bearing on the camera movement.

The general thought is if the pattern that you are tracking does not change over time, you should be able to track with auto-reference OFF as the pattern is always matched to the original reference frame. Let’s say frame 1 for clarity. So as the tracker advances through the frames, it will always check frame 1 for a matching pattern within the tracker box. If it matches, you get a keyframe at that frame.

If the pattern does change over time, auto-reference on is your friend because it checks the previous frame as the last reference instead of the original frame 1 for the pattern. This can help you get over difficult tracks, but the side effects are that you’ll potentially get more slippage because the reference is updating on each frame.

I don’t think stabilizer needs a rework “as it does what it says on the tin” however there could be multiple image and colour space factors that could cause it to fail. The suggestions that have been made so far are great and I would always be up for trying what works for you. As always, there is room for potential improvement and that can be requested via the Flame Feedback Portal.

I’ll climb off my soapbox now :blush:

Hi Grant, I think your missing the bigger point I am trying to make here and I totally get what you are saying and its totally valid but maybe I can show you since this screen record I tried isn’t totally clear, the issue I am having isn’t with grain as I did a test on color bars and with auto update reference on the track drifts which one would think if it works like you describe should be technically more accurate not less and theoretically not drift, if I turn off lets call it “AUR” it sticks perfectly, Moreover, if I say repo that bar while keep AUR on and apply that to a different batchfx and then just offset my reference the first frame always pops, it happens pretty regularly in shots I am working on where I have say ten of the same looking shots and copy that script across and just go into the tracker it pops. Again I am happy to discuss or screen share with you all since this might be bug behavior but not totally sure.

This happens to me and i find that if change the clourspace tab from auto to other it normally fixes the issue.

Hi William,

I did a test on bars and the tracker slipped as you said with AUR. I did desaturate the bars and you’ll see that the contrast between the colours is fairly minimal. This would add to the tracker slipping. I created white dots all over a black frame and tracked the dots. In this case there was no slipping with the option on or off.

I don’t disagree that something might need a tweak and I would be interested in seeing a video screen capture of the issues you are having so that I can repo this and we can file it with the dev team.


another + to have the tracker consider saturation as a target in its algorithm I am sure its in there but how it weighs this it shouldn’t slip between a green bar and say a red bar , but I will forward you a video , I have a limit here of 8mb so do you have an email I can send you a link?