We were recently confronted with what seems like a simple task: make a shot that had been previously tracked in Syntheyes longer, and provide a new camera that will retain the same basic camera info (fov, coordinates, etc) with the additional frames so that artists can simply drop a new camera into their action setups and carry on working.
Problem is, there’s a lot of talk in the syntheyes documentation about doing this, but there isn’t much there or elsewhere about HOW. And as we grow more dependent on 3D tracks, we would like to be able to do this more reliably. The current solution involves locking points and re-solving, which nearly works, but isn’t precise enough.
So, simple question: what are your best practices for extending a solve while preserving existing constraints?
I kinda suspect, after the voluminous but largely fruitless research I’ve done over the last few days, that there are two possible answers: 1) duh, a longer track will lead to a new solve, work in a way where this won’t bone you, or 2) duh, there’s a button that does this and there are no tutorials or manual pages about it because it’s so trivially easy.
But I hope that there’s a third option that details a procedure we can follow (or at least a pattern of words we can search that will take us to the relevant tutorials). Syntheyes seems to have a way to stitch multiple camera tracks together, designed for long, problematic scenes, that seems promising, but I haven’t had a chance to dive into that yet.
@julik@finnjaeger@inti.martinez ? You folks seem to know your way around a camera tracker. Any experience with this kind of thing?
I have never used sytheyes but if your track is locked to geometry like a lidar scan or something else redoing the camera from scratch shouldnt be this hard, Ive never attempted to add frames to a track but rather re-do and it will still be in the same world space as the old camera.
The it should just be a drop in replacement into any CG software. Oh and in 3De you could even wothout geo reassign 2D points to the old 3D points.
Yeah, that has been my general experience, but I also never used to bother with constraints or anything (because I still don’t really know how to use syntheyes properly).
This time we took the trouble to set up the orientation and ground and everything else, so when we re-solve with the new frames, that all gets re-set or, at the very least, updated slightly.
Hey Kirk, I don’t know if you check these videos out but the dude (Calvin, I think) puts a lot of stuff out and is very helpful, I feel like he’s be someone who might be able to give you the “duh” answer you’re looking for if you messaged him. TRACK VFX - YouTube
As someone who is also not a syntheyes power user by any means, I feel like this might be something (retaining the coordinate system and ground plane you set up) with the seed and lock functions in syntheyes, but again, I don’t reaaallly know how that works because I don’t use it.
Thanks @BrittCiampa! I’d found his tutorial about extending the camera path and it was very helpful indeed! Hadn’t considered messaging him though. Good idea!
We were collectively working up the courage to email Russ, but this might be a good intermediate step.
I feel like Russ would totally just be like “this is how to do it- and here’s a video of me doing it!” That’s just the vibe I get from him, he just really loves geeking out on 3d tracking, but I could be wrong haha
There is also a syntheyes users discord channel, I think Kalvin is a part of that, as is Matt Merkovich (who is a syntheyes genius/ guru and a super friendly and helpful dude).
Well, everyone can finally breathe a deep sigh of relief, because it turns out that, in most cases, extending a shot in Syntheyes is actually fairly straightforward!
And the trick, if you can call it that, isn’t documented as the key to extending your track because it’s so fundamental to the correct operation of the software that it hardly rates mentioning anywhere other than the literally dozens of places it’s mentioned (usually in the context of “and then you do this… and don’t ever forget to not do that”).
That thing is switching to “refine” which you should apparently do, and not leave the solver on “automatic” like an animal (or me, prior to three or four hours ago).
Beyond that, assuming you’ve set up constraints and coordinates to your satisfaction and to that of the software, simply replacing your footage with a longer version of the shot, adding blips or supervised trackers (or extending the supervised trackers you’ve already got) and then re-solving will largely preserve your scene geography.
Occasionally your world size will double (don’t know why, but I just set it back to the original value and carry on) and some locators will move around a bit to reflect a new solve, but anything you’ve locked in will remain in the same place and, crucially, your camera will not spin around to a totally new location/orientation.
Also: It seems that adding survey images and linked constraints and such will kind of help, but is such a pain in the ass that I don’t really see myself doing it much going forward.
Victor Wolansky’s FXGuide series on Syntheyes is what finally cracked this for me, btw. Give it a watch if you’re ever stuck somewhere. There are some great fundamentals in there that a lot of the more granular tutorials kind of gloss over.