How much skill level would it take to character rig a drummer in action?

[Steve Bentley]

As a fellow projection mapper, I guess it begs the question how real do they need to look.? Even with the best techs in hollywood we’re only now just seeing the other side of the “uncanny valley” (albeit still a hazy distance away).

Will you be animating them yourself or using motion capture? Then there’s the rigging, and painting of the weights – this alone can take a quite a while even in practiced hands. Textures too but if you go stylized that might be the easiest part of the job. People are just hard- we’re all so intimate with them we know when something’s not right: motion, the way the skin transmits light, eye tracking… when it’s not right it’s kinda creepy.

I’m not trying to discourage but painting weights and rigging can be an art and I find that depending on the action planned for the character each sequence may need special treatment depending on what they are doing – 3D characters aren’t quite as flexible as we rubbery humans are – at least and still have the character look good: without arms passing through other body parts, or shirts smearing as the arm passes by the torso, or underarms caving in. Shoulder joints are probably the toughest (and I think your drummer will be utilizing that spot a fair bit).

Here’s a thought though – what about just shooting the perfomers (you might have to anyway to get good motion capture data). Then map the real footage into a 3D enviro and do your warp as needed to fit your achitecture. A one day shoot could save you weeks. In the end it’s all 2D and an illusion, so whether it’s 3D animated characters or 2D footage, it all becomes a flat image plane(s).

[Hilary Tsai]

Thank you for your answer, Steve

I would consider motion-capture shooting, except that this show is a tribute to the band, Queen, and the models are supposed to vaguely resemble the band members (they don’t need to look realistic, I just need their iconic features to be apparent. Otherwise, they can have impressions of faces rather than real features, etc.) I would also be adding effects like x particles to play with their image and distort it. I was thinking of downloading a basic pre-modeled figure and customizing it to have the right look–then adding the instrument. Been looking into Adobe Fuse and Mixamo in addition to C4D. Do you think that modifying a basic figure is a better option than modeling from scratch?

[Steve Bentley]

I wasn’t going to say anything but since you brought it up…
We just went through hell because of Mixamo/Fuze/Adobe.

Adobe bought what was a sweet little company we have used successfully in the past. It was great: you could combine mocap data in the app and then export as a seamless peice-together of the data without the hassle of appending or retargeting or realigning the adjoining datasets. The characters all had the same bone structure… I could go on an on.
But now since V2.0 under adobe’s watch, they have broken it and see no point in fixing things since they are working on project Felix (project Pinnocio might have been a better name – they can’t even get that right). In the forums you can find a huge number of people asking what happened and where this feature or that has gone. Only to be told they are no longer developing and are now focused on Felix, and besides”less than one percent of users used that feature”. That “feature” in question was the combining of data in the app – it was the killer app, so I highly doubt that number. Its all very much a flip of the finger to long time users.

If you down load the example character (there used to be more of these but now just the one and it’s not even a single mesh! (So it hides a variety of sins that a single mesh wouldnt) and start playing, you will start to go “hey this could work”. Then the fangs set in. If you pick another character – different bone set. if you pick another mocap routine – different hierarchy. They have made a mess of what used to be an elegant solution for people who don’t do this everyday.
And fuze is just so much vapour ware right now. And because it’s cloud based I think if Adobe decides to scuttle it as they have done with Mixamo you’re stuck.
But not wanting to be a nay sayer only. Here’s a link that we’ve found helpful but has it’s own issues.
https://www.proanimationbank.com
This is just data – we’ve never gotten any useful meshes here even though they have some for sale. If you read their bio they say they scrutinize all data meticulously – can’t be true- because even in the previews you can see glitches and our sets are full of them. Women and men data sets have different hierarchies, and depending on the age of the data it may have another hierarchy again. The “this data links with Data set A” feature isn’t always true. So “Man walks to chair – man sits down- man types” – all are supposed to link together. They do but not without some major issues. Walk routines tend to be static in place while the linking “turns” tends to be in motion, so again, not so easy to link for some one who doesn’t do this everyday.
The up side – the data sets are cheap. $4-18. But very segmented to up their profits. They could be overly segmented because they can’t get the data to mesh seamlessly either.

There is a free data set out there on the web originally captured at Carnegie Melon U (sp?). You often have to go through a number of servers as it’s so old. The catch? Its quite noisy and uses what I would consider an old hierarchy style.

I think the overarching problem you will encounter with bought characters or motion sets is due to no set standard with bone hierarchies and joint directions. The reason we don’ t have a standard yet is that each use is unique and someone sitting and typing will have different heirarchy needs than a character from Seal Team Six. I think too that the investment to set up a system to do this in house means that you won’t find a lot of affordable boutiques that do this, and in the in house types are going to their bone structures for what works for their projects.
Within an animation with lots of different moves, you will often find a number of different versions of the same character depending on what he’s doing – and the closer you get to a proper realistic human mesh (especially with clothing) the more often that happens. Woody from Toy story most likely has one rigged character for most of the movie(I don’t know that, but given his stylized shape it’s possible). But other complex humans in other productions often have a new rigged version of the character for each specific task so that they don’t get hollow armpits or twisting forearms or flat butts when they sit.

What about shooting and using silhouettes – that way you can simulate brian may’s hair with a wig and in silhouette people will assume it’s him, whereas a model of him just won’t get there without a huge amount of work. We’ve done tons of this kind of stuff with mocap (silhouettes I mean) its very forgiving from a rigging point of view. But if you have the means nothing says real like real. And you get exactly what you want – purchased mocap is always a compromise.
FYI – For some reason this kind of thing looks better with a little slow mo on it.

[Hilary Tsai]

Hey Steve:) Your suggestion made me so happy about recording a guy with a Brian wig on. I actually might do that… Up until now I’ve been painstakingly rotoscoping them out of the poor quality concert footage and applying curves to get that silhouette look. I will definitely not be wasting time with Mixamo or Fuze now that I know how lame Adobe made them. Currently taking a look at proanimantion bank and watching tutorials on 3d modeling. I suspect I will end up modeling only the drummer and just using work-arounds for the others. Thank you so much for your incredibly helpful comments!

[Steve Bentley]

Here’s one we did with Mixamo data sets and our characters. All silhouettes but with the lights and atmo and a viewers own experience at concerts you don’t even think about the fact that the characters have no detail. Only the characters are 3D (at least outside of AE) all the other stuff including the earth balls we had bouncing around were done in Video Copilot’s element 3D and Optical flares in AE.
And with so much movement going on with the crowd grooving to the tunes your brain says it can see more than it actually can. Most of vision is memory, not whats right in front of us.

11553_concertanimation.jpg.zip