Motionbuilder and our official position.

We would like to use this chance to explain the way we test products to recommend them to our readers and what we see in MotionBuilder.

First: we are not paid to review anything. We honestly have to use the gizmo or program in close connection with the makers in our own projects. We don't do the "get the box, read the box, send the review" kind of writing that professional reviewers sometims do. For a product to be reviewed by a COW host, the thing has to have been in use in the host'
s studio for enough time to complete a real-life project. As we progress through the project we take note of the behavior of the reviewed software of machine. After the project is done --or a few of them in some cases-- we sit down and write our impressions on the whole enchilada... if it worked.

This means that we are open to all the suggestios from the makers and specialized users of any of the reviewed tools, but also that no matter the prestige of the maker, we will only review a product that actually performs. You will not read in the COW reviews of the "Hatchet Job" kind, simply because a tool that consistently under-performs will NOT get a review.

Another good thing: we don't own the programs or tools we talk about. Sure, you are a Maya geek like me, you get to host a Maya forum, but you don't get to review Maya. That way we don't have the emotional and financial investment talking, and you don't get the bias of the "I bought it, it must be the best" kind. Any enthusiasm you DO get comes from the tool and its performance in OUR projects. Whether we have a service shop or a development one, we can only sing praises of what DOES contribute and improve the putting bread on the table we need in order to live.

After five months of tests with MB, we were unable to integrate it reasonably in any of our pipelines. We did several animations using it -- Mr. Wright and Brad are aware that is easy with MB-- but NONE of them could be imported back into Maya, even with the invaluable assitence of the Alias MB crew. It turns out that MB only deals with complete models, and anything made out of parts is outside its capabilities. We use "singleskins" only for some tasks. All the models were transformed by MB into an array of overlapping yellow spheres: the articulation icons. The bones and the models were not to be find. They did NOT translate back. That situation was confirmed to complete satisfaction --or lack thereof-- by the team assigned to the solution of the problem at Alias. It ALSO turns out that the fbx translator plugin is not wholly compatible with Maya 6 OR 6.5, so during the period of testing the fbx was actually re-written. After failing to import the models into Maya anyway, we had to forego further experimentation as the project had to be delivered. We switched our Maya full blast and crammed until we recovered the lost time and delivered the gig reasonably on time, only in February rather than December, thanks to MB's lack of performance. We still tested MB until March 2005, so our take on it is not a hurried one.

We also ascertained to complete satisfaction that the facial animation system only works in Maya for polygonal faces that do not blend, up to a point. As with the bodies, you need specially built faces and bodies to use MB, and have to depend completely on it, so forget about using the results and fine-tuning them in Maya. If you build a face as MB likes it then it is up to chance whether your animation will travel satisfactorily back to Maya. It also means that ANY CHANGES have to be done in MB and re-imported, re-aligned and re-blended. Frankly too much work if all you need is to emote a bit better the dialogue.

We think MotionBuilder is a clever re-purposing of an old tool. A great idea, not yet completely realized. We will hold the review until it is ready. The FilmBox from Kaydara is still there, too much up front and personal. The tool has changed little since we used it to clean all that motioncapture for The Matrix and the subsequent games using a thirty-two camera Vicon. The interface is still cluttered and even tho I was able to navigate it with ease in a few hours, nobody at the shop felt it behaves or handles as an artist's tool, we all agree the thing feels like a Technical Director's rig, exactly as it used to feel but with a few prettified dialogs added. It lacks elegance in everyone's opinion, for whatever THAT is worth. The name-based import system is frankly ridiculous for a professiona application of more than US$100 in cost.

BTW- we used that names-scheme in the old days of wavefront. That is relatively easy, if clumsy, but the only reason it is there is because trhe Vicn NAMES the nodes itself. There was no way to edit them at the beguinning until AFTER the data was in the computer.

The way this should be done is by parsing the bones, Exactly as any application made for animation does, be that Maya, Max, Lightwave or Softimage, both S and XSI. dependaency in something SO fragile as USER NAMES is a serious design mistake. And compatibility has nothing to do here: You give it an fbx file anyway, so why can't it parse the bones as it should?

Anyway. We think that MotionBuilder is a great idea on the way, but calling the glorified FilmBox an "animation SYSTEM" is a bit of a stretch. We cannot recommend our readers to take the cost of two Maya Completes and sink it in an app that works for easy stuff but fails as soon as things get demanding. There is still that old plugin "skeletonworks" that does pretty much the gist of what MB should do and used to be freeware. It only does bipeds but it does them just as well. If what you want is not to have the aggravation of building the skeleton, that works too.

And MB offers only SOME bipeds and SOME quadrupeds for US$6K. Reverse-jointed animals like Ostriches need not apply. And the animation controls are NOT even selective. You make a keyframe, it happens for ALL the joints around. Not usable for fine motion, since you end out keying curves you'd need to let flow. It is made for easy REPAIR of mangled Motion Captures that's what, and as the readers who use it have pointed out, it is very easy to use. Good luck reading back your stuff, tho. Some will, some will not.

If you don't need anything sophisticated, using MB is a lazy, real easy way to do it for only US$6K. A note here: in Softimage, a properly rigged skeleton WALKS if you pull the anchor-pelvis forward. THAT is what kinematics are supposed to do. If you build it (right) it will walk, as it were :) Where is that in MB? Nowhere, because it is NOT an animation package, it is a motion package with handles grafted onto the end effectors. Yes, you can animate with it, but it's really meant for corrections.

We do respect the opinion of our readers, and as we point out, the thing is usable. However, we have dedicated our life to animation and we don't think this tool is ready. We have been using animation software for the better part of 20 years. We are far from hidebound and our shop sees at least one or two new programs a month, given the variety of projects I get into: Live action, animation, abstract animation, compositing, recording, motion control, motion extraction and matchin, shading, render, automatic animation, hardware... We are relatively used to write our own tools for our projects. We use our own render queue, and any script coming into the shop for bidding gets passed through our own SUSHI app, to extract the elements and the staging grids. When we say that the interface of MB is raw, we don't try to disparage the efforts of any user, just to evaluate the product from the same perspective we do all tools.

Even so, we just emit a recommendation either way. Each reader has to decide if the remaining virtues of a particular tool are something needed in a project and worth investing on.

take care



.k.