Cineversity

I’ve followed every youtube tut and post here but can’t get this to work.
I can get a Constraint tag to work with a standard model, but I’m working with a .OBJ sequence of a face that I converted to an Alembic sequence. I have a male bust that I want to match the movement to the face. Every way I’ve tried, it doesn’t work. I can get some movement but never anything that looks properly aligned to the face.

Or maybe theirs a totally different way to match these two objects.
Any suggestions would be greatly appreciated.

Also, I included a link to the file in case it helps. Thanks,
https://www.dropbox.com/t/tPX8I25rEFjUU9RC

Hi mohamed.kadva,

Thanks for the data and for using DropBox, very much appreciated.

To be clear, that is not an issue of any constrain, options, nor functions.
There is not a single point on this mask that is stable: All is moving. With moving, I have in mind that the overall movement has not a single point that matches it.
To clamp anything to a single point will not change the rotation; it will move things (but also not scale it).

Please have a look at the file and the preview:
https://www.amazon.com/clouddrive/share/4C1FSiyPY1nVxqGM35HknO12sJLt2tlT8CfLbXUImyz
I did not include the abc file. You need to place this file into the folder you shared.

I took three points from the forehead, which are typically the most stable, if at all. One for the position, one as a target, and one as an up-vector. That is the most stable transfer typically, but if all points are independently moved from each other, that is NOT a way to go. But it would be the closest at all, AFAIK.

=====

The main question would be, where do you get this face-tracking from? Moves by Maxon? If so, then you have the main movement data. You need to get the original data. From there, it is easy. It is given and can be applied. Let me know.

There were more discussions about this
https://www.cineversity.com/forums/viewthread/3961/

All the best

Sorry about the delay in responding. Thanks so much for the help. It worked perfectly when I updated with my Alembic sequence.

I’m trying to replicate this from scratch with the expresso editor and new points on another face model and having issues. Is there a tutorial on cineversity or somewhere that comes to mind that I could follow along with? Or what specifically would the search term be that I would be searching for?

For additional context that might help, this isn’t from any mocap like moves but a machine learning model, so that’s why the results are so jumpy.

Thanks again,

Thanks for the update, mohamed.kadva

Think of Xpresso as a language, it is not needed to know all [words] from the start to “xpress” something. If there is a question, just ask here.

Typically the main thing to know is the information that is inside of the wires between the nodes. (Real, Integer, Vector, Matrix, etc.) Also, here, that will feel natural to you after a while of working with it. If you hover with the mouse over an input- or output-port, you will find what information is provided or expected in the lower-left corner.

There might be a tutorial somewhere in the world that discusses exactly this set-up (from above), but each solution with Xpresso is kind of tailored to your needs.
It follows an information flow. What object can deliver what, how should that be changed, and where will it be used on end.
What information do you have, is there any need to change it, and what part is needed somewhere else?

What do you have? The face mesh. It has too many points, so how to get a few of them, and how to get the index of that point? To get an idea of how that works, please have a look at the screen capture.
https://www.amazon.com/clouddrive/share/eTva1vA624UQDNrJWQ10qX8jBkKj7SNOR2TCeobg8LJ

Let me know if there is any question. I’m happy to look into it.

All the best

P.S.:
I did a short series about XPresso.
It is around 15 years old, but the system hasn’t changed a lot, my english was quite rusty back then, sorry about that.
https://www.cineversity.com/vidplaylist/sassis_tootime_tips_xpresso/