Collaboration between animators working with XSI

[sonictk]

Hi guys:

Usually I work alone, but for a animation project, me and another guy are thinking of doing a team animation. However, as I have no experience in working with XSI on a collaboration basis, I'd like to ask:

- What best practices are there for collaborating on an animation with multiple characters interacting with each other in a single scene? How would two people (or more) be able to work on the same scene at the same time, assuming we had to be able to view and animate each other's characters at the same time, along with other environment assets that would be constantly updated?

- Would we use action sources/referenced models (i.e. each of us works on our own referenced model and exports its animation as an action source, which is then referenced into a 3rd scene file with the environment and lighting that would act as the 'final' scene to be rendered. Right now this is what I have in mind, since it would allow us to work in tandem with each other using version control instead of having to wait for one person to finish their edits to the scene before passing it to the other person. (Do referenced models store action sources as well as animation?) Also, if we were using referenced models, would we be able to say, animate that model with an environment, but leave out the environment when updating the model automatically without having to go in and delete the environment each time? I have very little idea as to how team production in XSI should be done, sorry if this sounds really dumb.

- Alternatively, would we be able to animate seperate .scn files, each with our own character in them, and merge them at the end for the final render scene? The main drawback here that I can imagine would be that we have no idea of how our animations look together until the very end, and then we would have the same issue of having only one person being able to work on the scene at a time again.

Anyone who uses XSI in a production or a team-based environment, please lend your expertise! I would very much appreciate any advice that anyone has to offer for such scenarios, as well as common mishaps that you have faced so that I would know to avoid such issues

[Draise]

I've done it a little, but not to your extent.

But things I've done: used a dropbox, skydrive, googledrive option with referenced models and a mutual project structure on the cloud so one person could work on a scene, save, then the other could work on the same scene (first one on layout, second on animation, third on lighting and render). Sometimes I would export the model and they would load it when needed into the final scene to be rendered or animated with other characters, with the animated model with version numbers, or shot numbers. Merging scenes didn't occur to me, but having stand in models with already set animation (action clips are stored with the *.emdl) did work just fine, like a charm.

But due to not having a very good compatibility with multiusers on a tech side of things.. a good work around is good planning, having the same audio file to work with (prior), extensive storyboards and even animatics with storyboard shots and sound or even animatics with rough animation that is reviewed and pre-thought together online. Once you both have gone through the idea of how it should be, the acting and animation will be easier between you all.

Just.. some ideas.

[sonictk]

Hey, thank you so much for sharing your experience! I really appreciate it, knowing that at least it's possible if nothing else

Could I ask though, since you mentioned that you used referenced models: I'm imagining you had something similar to this...?

Code: Select all

Master Scene file (.scn) (to be loaded, references updated, and then rendered)
    --> Character 1 geometry with animation  (referenced .emdl)                  --< Worked on by Team member 1 
    --> Character 2 geometry with animation (referenced .emdl)                   --< Worked on by Team member 2
    --> Environment geometry with lighting and shaders (referenced .emdl or FBX)  --< Worked on by Team member 3
    ---> etc...

However, you mentioned that:

then the other could work on the same scene (first one on layout, second on animation, third on lighting and render)

I'm not sure I understand this part fully. If the character animations and environment were being worked on as separate referenced files, why would you guys have had to work on the Master Scene file at the same time? Wouldn't each member work on their own section seperately, update their version of the other team members' files as needed, and compile into the final Master Scene .scn at the very end for rendering? (I'm not trying to be snarky, I'm genuinely curious as to what problems such a workflow might have)

And thank you for the advice about storyboarding/animatics! Both of us fully intend to work closely together on this, and as we are planning to use SVN for asset version control hopefully most of this pain will be lessened as far as getting the same Softimage project with all the necessary assets (audio, supporting planning resources such as animatics etc.) across to all team members is concerned.

EDIT: Another thing I was wondering about...will .emdls/ saved out from different versions of Softimage play nice with each other? I ask because I have 2013 SP1 at home...but my other team members will have to work with 2013 vanilla since our school refuses to install any service packs for the Autodesk suite. >_< (I know .scns sometimes have odd issues where animation is concerned, which is what worries me too)

[Draise]

Yeah, I was kinda ranting.

Due to different time schedules, we didn't have to use the referenced models in the end. One worked in the scene, then saved. I opened later, worked, saved over, and so forth; later in layout, we used the main server here and the renderfarm and the render scenes on local networks to later get the final image sorted with the animation.

Due to the final render size of each shot, and due to saving space on the cloud (and bandwidth) we didn't have the final render elements/scenery online. Also the render passes system/file system was complicated to preserve with large changes in the final render layout with ever changing lighting and artistic desires of the creative director. So the cloud only had a scenery reference scene depending on the shots (with reference props only), default characters and the final animated characters (references imported to render) and animatic based material and shotlists and tasklists (and audio, very important). I would import the animated characters into a render scene locally for the farm later.

Heh, maybe I shouldn't give all my secrets away! Or.. maybe if you want, I can try work with you guys a little here and there if I have the time. =)

-Cheers

[sonictk]

Due to the final render size of each shot, and due to saving space on the cloud (and bandwidth) we didn't have the final render elements/scenery online.

That's one thing I'm really worried about, especially since XSI files are all binary and SVN doesn't handle that very well. =\ I'm not sure how we're going to work around this yet, though I'm asking my programmer friends how they deal with version control for such things. So far most people just seem to live with it.

If nothing else, we might have to end up with a similar workflow to yours: where essentially one person works on the scene at a time. Hopefully I have enough time to test an alternate workflow that works I would hate to be essentially biting my nails while waiting for someone else to finish their work on a scene and hope everything still works...

Also the render passes system/file system was complicated to preserve with large changes in the final render layout with ever changing lighting and artistic desires of the creative director.

Could you elaborate if possible (or convenient? ) Complicated because of the sheer size of the scene with multiple passes to be rendered? Was the shader structure not being preserved (or some other technical issue) or was it just more efficient to change the lighting on the spot as your director saw fit instead of constantly updating the scene with the latest version from the cloud?

Heh, maybe I shouldn't give all my secrets away! Or.. maybe if you want, I can try work with you guys a little here and there if I have the time. =)

Thank you so much! We're actually thinking of entering the AREA contest (http://area.autodesk.com/contest/faq#content) since I've tried all the rigs, and the XSI ones seem to be most animator-friendly (Maya's one makes use of HIK and that brings its own issues with it, such as synchronized Quarternion curves by default, etc.) so maybe you'll see some WIP stuff from us in a few weeks! Right now I'm just focusing on getting a working workflow (heh) so that I can focus on bringing my other team member up to speed in XSI without any problems (Our school only teaches 3ds max/Maya, and only teaches students how to animate in Maya)

EDIT: A quick question: Could simulations (ICE particles, Lagoa Cloth etc.) be referenced into models as well? Or would they also have to be kept within the same scene? I tried looking at the XSI documentation but there wasn't much mention about this (Any technical issues that you've experienced with such things would also be welcome)

Again, thank you for your advice!

[Draise]

If nothing else, we might have to end up with a similar workflow to yours: where essentially one person works on the scene at a time. Hopefully I have enough time to test an alternate workflow that works I would hate to be essentially biting my nails while waiting for someone else to finish their work on a scene and hope everything still works...

Well, depends how much time you have, just try not to panic and get angry at your members like some needy wife.

Also the render passes system/file system was complicated to preserve with large changes in the final render layout with ever changing lighting and artistic desires of the creative director.

Could you elaborate if possible (or convenient? ) Complicated because of the sheer size of the scene with multiple passes to be rendered? Was the shader structure not being preserved (or some other technical issue) or was it just more efficient to change the lighting on the spot as your director saw fit instead of constantly updating the scene with the latest version from the cloud?

Yes, no, yes and yes. Shaders preserved ok, just had to make sure the texture paths were relative. Scenery and pass information of a scene file were huge, with scene files up to <150 MB. Which takes forever to sync over the internet. So exporting a model at <15mg was a more viable way to do things, faster. Also changing lighting in the final scene was much easier with everything done and ready locally, again, updating the default render scene on the cloud would mean having to reupload the <100mg file again (or more or less 5-20MG file with the referenced scenery with a large sum filesize to sync if you change everything).

I also got annoyed with reference models, because sometimes in loading a referenced model (large ones or high poly ones), it may crash SI. It is best to make a default pass, hide the partition, and load the reference there (hidden). I think it has something to do with the viewport code or my video card not being able to update reference models well once updated.

Heh, maybe I shouldn't give all my secrets away! Or.. maybe if you want, I can try work with you guys a little here and there if I have the time. =)

Thank you so much! We're actually thinking of entering the AREA contest (http://area.autodesk.com/contest/faq#content) since I've tried all the rigs, and the XSI ones seem to be most animator-friendly (Maya's one makes use of HIK and that brings its own issues with it, such as synchronized Quarternion curves by default, etc.) so maybe you'll see some WIP stuff from us in a few weeks! Right now I'm just focusing on getting a working workflow (heh) so that I can focus on bringing my other team member up to speed in XSI without any problems (Our school only teaches 3ds max/Maya, and only teaches students how to animate in Maya)

Cool. Sucks I'm not north American. And yes, Softimage seems to not be taught a lot. But I do prefer using the animation and rigs from SI over Maya. I had a colleague who suffered in Maya, meanwhile I was pumping out 4 times his quota in SI with less training. Looking forward to your WIPs, I love following WIP's.

EDIT: A quick question: Could simulations (ICE particles, Lagoa Cloth etc.) be referenced into models as well? Or would they also have to be kept within the same scene? I tried looking at the XSI documentation but there wasn't much mention about this (Any technical issues that you've experienced with such things would also be welcome)

Not sure, haven't tried. I would assume if you write the simulation cache, (and maybe import the model itself,) it should work out ok (maybe.... as a reference, no.) I find that writing the simulation cache/freeze animation of your cloth or particle simulation really helps in the renderfarm and/or online collabs (network related work).

Again, thank you for your advice!

No worries. I'm going to bed now, peace and good luck!