PDA

View Full Version : Can these 3D techniques be done on indie level?



FilmBoy77
04-02-2014, 11:12 PM
Obviously the example below is from a $180 million dollar budget movie, but can the process of the facial and body movements be done using software like Maya, 3D studio max and ZBrush?

I'm interested in knowing how you can get realistic animation from CGI characters. Not because I want to do it myself, but I want to understand it enough for my own benefit.

Thoughts?


https://www.youtube.com/watch?v=9Vsxq6Afk3o


https://www.youtube.com/watch?v=8MWkYyr8hw0

Derek Chingwell
04-04-2014, 08:12 AM
Weta's pipeline uses Maya, Mudbox, ZBrush, Motionbuilder and many (many) other CG tools...

It's possible to recreate animation like this if you're skilled in the software and have experience (and talent)... but bear in mind that you're seeing the work of teams full of TD's, animators, Leads and Supers.

Mocap (full body) is often a base, that is then use to layer animation on top (using motionbuilder) and this is often sent to custom rigs in Maya, to be re-animated again... it all depends on the pipeline. Facial tends to be data driving blendshapes (morphs) in a movie pipeline. Basically a TD or team of TD's create a facial rig that is driven by the base capture data and then augmented/re-animated and tweaked by hand... it's very (very) rare to expect clean (un-edited) data to make it into a final shot.

If you have a good rig (full body and facial) you can create great key-frame animation, but you'll find that mo-cap data ends to be better for 'realistic' humans as the subtle nuances of human motion are picked up (multiple small rotations on bones and positions for example) that are often not added when you're keyframing... that's why mo-cap 'idle' animations (characters just doing nothing) tend to be better than key-framed idles (no matter how good the animator is).

Of course, once you're out of the 'motion' pipe, you'll be into the lighting/surfacing and render pipe. Here you can have bump/normal maps driven by morphs (so a morph can close an eye-lid and that closed position will, in turn, drive a stress map that's bump/normal to add subtle skin wrinkles around the eyes etc). There might be a side step into simulation next (ie clothing or jiggling muscles and non key driven motion based elements) and then onto the many layers of texture/lighting that are based on each shot's needs... then this gets broken into layers and rendered and moves to compositing... where, once again, more work is done.

BTW, I've worked (off and on) in CGI since the late 80's (when it was very different to the world we have today) and it's really down to the size of the teams/budget involved and (importantly) the skills/talent of those involved. :)

nosys70
04-04-2014, 05:42 PM
this is very possible with low budget, but you will be a one man team and expect to spend a lot of effort to make all pieces fit together. Today you can digitize a full character for almost free with a simple game accessory (the Kinect camera or similar). Then using softwares like Faceshift or Dynamixyz you can animate facial feature, or even full body. there are tons of free utilities allowing scanning, animating, mesh editing, motion capture.

FilmBoy77
04-05-2014, 08:19 PM
This is good to know.

What's the process in which these cool CGI characters/monsters get made?

It starts with a sketch of the creature I assume, then who sees that next in the process? The person modeling? And then the texturing happens and then it gets rendered out?

Josh Bass
04-06-2014, 03:43 AM
I saw them do that bit where they used a kinect for a 3d body scan for a 3d printer model on an older episode of Big Bang Theory! I was wondering if that was tv bs or a real thing you could actually do. Their stuff is usually pretty well-researched, so i guess they win again.

nosys70
04-06-2014, 12:55 PM
there are many path you can use, the result will be always the same.
In big studio ,they usually take an artist view from drawing, then have them sculpted in 3D, then digitized, then rigged to standard skull (bones and joints) they can animate.
the only difference with a low cost or high end product will be the finesse of details and the effort spent to the model to fit the wishes of the producer.
there are also many details that must be managed separately , like hair, furs, texture etc... who can be a challenge.
If your character is a regular biped, you can skip many step, because there are lots of tools (for example the biped function in 3Dsmax) or product you can buy preanimated. You just need to provide your model. And you can easily create a model in 3D with a simple web cam today.

check https://www.mixamo.com/workflows, they have alot of information on the workflow.
that is for body, because body motion is very similar for all characters.
For faces, this is more difficult, particularly because you need to lipsync to the speech in the scenario so most company turn to real time motion capture, but you can use some converter able to convert audio file to animated lips with surprising accuracy.
basically you stick some trackers on somebody's face and make a link of these trackers to points on the computer generated face.
you can add apmplifier to some point, so a small move on the actor's face become big move on the character's face.
you can find some freeware for this.

But again as said previously there are many tools, with many input/output formats, so the first thing you need to find is a valid workflow between all these tools. 3DSMAX is a great tool to use as central point because it can use many plugins, handle many formats, and you can find many websites, forums online for help.

Derek Chingwell
04-06-2014, 03:32 PM
In terms of getting a character from concept to completion (on a decent budget production), there are quite a few steps.

The Art/Character team will take a brief from the Art Director (or team Super) who'll have been briefed by the Director. There'll then rough character sketches made (a massive variety of choices normally). These are then reduced down to a few versions the Art Director likes and these will be shown to the Project Director (and sometimes the Producer). Often, cost implications of each design will be discussed and at least someone from the Character team (normally a Super and a Rigging Lead or TD) will have an input.

When a final design is signed off (can be weeks or months - after lots of artwork is produced - work now is normally 2D paint based as most artists are comfortable in paint packages), a 'character' package is produced that includes turnarounds (views from all angles), texture ideas, colour keys and mood work (for an idea of the character in action).

Sometimes, clay Marquettes are produced, but this is normally to see how shapes work and translate to reality (albeit scaled down :) These are scanned sometimes, but the scans produce NURBS data and tend to be used as the basis for a completely new model (from sub-D's, as these are much better in a Renderman pipeline) as scans are way too heavy in data for rigging, layered simulations (ie flesh/muslce sim and skinning) and cloth sims. Rebuilding as scale-able sub-D means the asset is very light in the production pipeline until it's at the render stage. Maqs are good for PR/making of etc, but most TD's and Supers will avoid scanned data like teh plague, it's heavy and error prone, as you do not have control of the mesh.

So, the model is remade, then a version is rigged (normally Maya - but I've seen pipes based on Lightwave, XSI and Max in my experience and there's often a bi-directional rig built in Motionbuilder that's specifically designed for the mocap shoot). The Motionbuilder rig tends to be bones (direct mesh manipulation via rigid weighting) and morphs. This version will be used to see real-time mocap (Director feedback) and then be used as a processed data rig and sent with mo-cap clean up moves via FBX to the TD's who'll prepare it for keyframe or simulations (the complexity here depends on the gig you're on, the specific pipeline and the budget).

Normally, the character will have a layer of key-frame animation to augment the mo-cap (either in Motionbuilder or Maya etc), and this will then be sent up teh pipe when approved by the Animation Supervisor and Director, to simulations, who'll then look at adding anything from hair, fur, skin, clothing or 'secondary' driven animation such as flesh jiggles or muscles etc.

Going on at the same time, the texture/surface teams will be working on the textures required... and the Rendering team will be looking at breaking down scenes into elements that are required (complex or simple depending on the shot)... and of course, Lighting will be dealing with lots of decisions too.

I'll stop here, as you see, even before the motion is finished, and the character might not even be textured/lit or test rendered, it can already have been through many (many) artists, supers, leads etc

I can always post again with more stages in the pipeline... but this reply has ended up being pretty long - LOL!

FilmBoy77
04-06-2014, 10:38 PM
Thanks guys I've read what each of you have said several times because some of the terms I'm not familiar with :)

If I were interested in filming/producing a one minute and twenty-two second scene like the troll scene from Snow White and the Huntsmen, how much do you think a ballpark figure would be for just the CGI of the monster? On an indie level would the workflow pretty much be the same as you mentioned above?


https://www.youtube.com/watch?v=hMx7OP7yiHA

Derek Chingwell
04-07-2014, 07:02 AM
Thanks guys I've read what each of you have said several times because some of the terms I'm not familiar with :)

If I were interested in filming/producing a one minute and twenty-two second scene like the troll scene from Snow White and the Huntsmen, how much do you think a ballpark figure would be for just the CGI of the monster? On an indie level would the workflow pretty much be the same as you mentioned above?



Well, if I was quoting realistically and was at a big studio (Framestore or DNeg in the UK etc), I'd easily say around $3/4 million maybe even more, as I'd be looking to quote for the rest of the shots too (and bundle a package of shots to try and make a profit).

SW&TH was about a $170 million dollar project, realistically, I'd presume the CGI/VFX budget was around $40 at a guess... could be more as the Director was not a 'name', so the the studio was probably willing to put more into the VFX budget to protect their investment.

Before you think that's easy money, you need a fully functional CG studio, a renderfarm, software licences and LOTS of experienced 'core' crew as well as freelancers to pull off something like this. You have to consider salaries, normal running costs (ie non CG staff), overheads etc etc - and a shot like this would probably be at least 6 months (although I'd guess it was 9 or 10 months).

Now, break down the finished shot (you linked) - how much of that is CG? Well, the Troll as a 100% CGI element, but there's also a mixture of CG practical elements (objects the Troll interacts with CG trees etc), plus landscape, potential VFX for liquid/particles, sky work, and that's before we talk about the match-moving needed for the cameras (which is more difficult than normal as the gig was shot with anamorphic lenses, making the tracks harder), green screen elements (actors against backgrounds) and also actors being pulled (roto work) from the live action plates. Also potential CG digital duplicates (for mid to long shots of the actors with the troll) - and then we have on-set VFX supervision, plus recording of data (for lighting the CG later) and all this is before we're really getting into 'small' details.

As you can see, it's pretty complex - more so than most people realize.

A good sequence like this 'works' because it's well planned and has many elements that help the CGI work with the live action and integrate... that's the key part of the magic. No matter how good the Troll movies and looks, if it doesn't integrate or feel like it's 'there', the sequence fails.

Now, on the other hand, yes, it is possible to do on an indie level - but you really need to think on the following:

a) have to know what you are doing (and have en experienced team from DP to Director and CGI/VFX team who know how to pull this stuff off)
b) listen to your VFX Super who'll help in shot choices (look at Asylum movies for awful shot choices and poor CG to see what I mean)
c) work within limitations of available (off the shelf) software as you can't afford to get custom code written to help you out if there are problems
d) always be flexible and be prepared to compromise shots and change things so they work... you can't be locked into a 'perfect' vision when doing this at an indie/low level.

To understand or put things in context - look at Game of Thrones... it as a massive VFX budget for a TV programme and is at the top end of TV budgets... BUT, there's nothing that even approaches the complexity of the sequence you've linked. It just isn't possible to pull off on a TV budget (even like GOT), as a TV series is turned around in 9/10 months and will spend way less on VFX/CG.

I mention this as the way you should think on an indie level should be Game of Thrones as the pinnacle of what you would be looking to pull off... and Asylum stuff being the worst ;)

nosys70
04-07-2014, 04:05 PM
the sample movie you gave is very fast paced and motion blur help a lot to hide bad details.
unfortunately the scene shows an heavy interaction of the characters with the environment and this is very diffuclt without careful planning, because
you can expect that must of the environment will need to be created in computer too.
if you go the scan route, you can scan a mid sized figurine, about 2 feet (using a 3D scanner or simple pictures) to get a volume made of a cloud of point.
Then you need to turn this cloud of point into a mesh (many free utilities allows this).
The problem of this technique is you create a random mesh, the consequence being the model is one big surface and this surface will distort without constraint. The result could look very strange on some movement.
To avoid this you need to reassign a new mesh that can follow some guidelines (like muscles, bones, body parts or materials)m this is called retopology.
A nice tool for this is topogun.
many cheap 3d creation skip on details by using a trick used for games: texture map.
you can have a smooth model, and apply textures to make it look rough.
this is used a lot to skip on body hair, but it looks really fake on close-up.
there is not one tool to make all, so many company are working by layer, adding each layer with a specific tool.
for example you start with a basic character , smooth, grey, with no details, just to test scale, animation, placement.

then you will add some basic details, especially if your character is made from both soft hand hard body parts that will require different physics.
finally you will add all the layers for color rendering, transparency, reflection etc...

look for product like scenect, skanect, kscan3d , 123D catch, meshmixer, faceshift, mixamo, topogun, mudbox

this guy has created a digital copy of himself, it is amazing
http://forums.cgsociety.org/showthread.php?t=1019262


for video, the job is a lot more easier du the the relative low resolution and the motion blur.
If you need real close up, you can build 2 models, one full size, and another (just he head for example) with higher details resolution.

FilmBoy77
04-28-2014, 07:10 PM
Great feedback guys. I found this short on vimeo where in the description they said they produced this for a "few thousand dollars". I notice the director also is credited for some of the modeling and animation so it looks like it's a situation where they did a majority of the CGI themselves which is why the cost is so low. What are your thoughts on how much work and time would be needed in regards to the 3d modeling and animation in this short film?


https://vimeo.com/71702428

Derek Chingwell
04-29-2014, 04:16 AM
Great feedback guys. I found this short on vimeo where in the description they said they produced this for a "few thousand dollars". I notice the director also is credited for some of the modeling and animation so it looks like it's a situation where they did a majority of the CGI themselves which is why the cost is so low. What are your thoughts on how much work and time would be needed in regards to the 3d modeling and animation in this short film?



Check out his CV and experience (D9/Thor etc) - that's why the VFX can be pulled off for a few K's, he knows what to do - and how to do it :)

Really, there's a ton of work here, if you know what you're doing, it can be done (from camera tracking/matchmoving through to all the CG elements). You'll note the variety of jobs in the credits.

This is different to the first set of stuff you linked... that was a 'living' character added into a scene (ie organic) and this one has robot characters (hard surface) added into scenes. Whilst they seem the same, it's a very different animation/rigging pipeline for a Mech/robot (etc).

If you're looking to aspire to this type of work, you need to think about some formal training, as you need a foundation in the basics - before you can jump in and start adding CG elements into your own movies...

In terms of time for the CGI/VFX, with a skilled (experienced crew) I'd say 3 months as a guess on timescale (if working full-time). If not, maybe 6-9 months as a part time project (it all depends how busy you are with other jobs).

FilmBoy77
04-29-2014, 07:37 AM
Check out his CV and experience (D9/Thor etc) - that's why the VFX can be pulled off for a few K's, he knows what to do - and how to do it :)

Really, there's a ton of work here, if you know what you're doing, it can be done (from camera tracking/matchmoving through to all the CG elements). You'll note the variety of jobs in the credits.

This is different to the first set of stuff you linked... that was a 'living' character added into a scene (ie organic) and this one has robot characters (hard surface) added into scenes. Whilst they seem the same, it's a very different animation/rigging pipeline for a Mech/robot (etc).

If you're looking to aspire to this type of work, you need to think about some formal training, as you need a foundation in the basics - before you can jump in and start adding CG elements into your own movies...

In terms of time for the CGI/VFX, with a skilled (experienced crew) I'd say 3 months as a guess on timescale (if working full-time). If not, maybe 6-9 months as a part time project (it all depends how busy you are with other jobs).

Makes sense! No I'm not trying to learn how to actually do this type of work, I just want to understand the mechanics behind it enough to be able to hire the right people who can do it :)

Just out of curiosity do you do any of this type of stuff? Or do you just know a lot about it in general?

Thanks!

Derek Chingwell
04-29-2014, 08:52 AM
Makes sense! No I'm not trying to learn how to actually do this type of work, I just want to understand the mechanics behind it enough to be able to hire the right people who can do it :)

Just out of curiosity do you do any of this type of stuff? Or do you just know a lot about it in general?

Thanks!

OK, makes sense now :)

You need to find a good VFX/CG Supervisor, he/she will be able to help you out a lot. A good CG/VFX Supe is the heart of most jobs.

To answer your curiosity question, yes, I do this type of stuff (as a job). I've worked at a few CG outfits in Pinewood Studios and Ealing Studios (UK) and been in the industry quite a few years (late 80's onward), so I've seen the growth of CG from using Symbolics and Cubicomps through the SGI dominance in the 90's and to the place we're currently at. My experience is split between CGI and film, the 2 industries I've worked in.

My advice (if budgets are low) is to see if there are any CG courses/University/Colleges near you location - and see if you can hook up with some students. They'll be heavily into all the tech side (the things you're trying to pick up) - and they might help out. Benefit for them will be working with people they can learn from in terms of film.

msorrels
04-29-2014, 09:10 AM
If you were to do this the traditional Hollywood way it would be a lot of work and fairly expensive. But you could do it more like TV VFX or Indie VFX and get the costs down a bit. Basically the difference is something like this:

In a Hollywood movie every VFX shot is designed and then implemented to that design. Someone has to make early versions of the 3d models, which then have to get approved, then modeled, then approved, then textured/surfaced, and approved, then rigged, and approved, then animated, and approved, then composited, and approved. This kind of overhead makes it very expensive when you are paying by the hour. Which of course is why producers always flat rate bid VFX and make the VFX company and artists eat the time.

A more indie movie/micro budget version though could go a bit differently. Namely you don't do nearly as much design, if any really. You also use stock 3d models as much as possible and kit bash to make things specific. This short includes a credit to Turbo Squid, a 3d model marketplace. There are several that can sell you a pre-made model, often times textured and rigged, but you give up that design control.

So for a short like this you buy a 3d robot model, like this one http://www.turbosquid.com/3d-models/3d-model-robot-bot/646376 (which is pretty close to what was in that short I think). You'd need to surface and texture it, but I have a huge library of already built surfaces and materials and could do that quicker than any Hollywood production. It's kind of like how indie films use the locations and sets they have access to, instead of building a set from scratch to match a design 100%. If you are flexible on what parts you use and use what you can get easily, you can get the time down on the design and model building phases. You still have to pay through the nose on the animating and lighting phases. And camera matching, rendering and compositing. But you start to get a project down to something a handful of VFX artists can do.

Something like this you'd need to break it down into VFX shots. Each shot's cost is a function of how complex the shot is and how long it is. There are a couple of city environment shots in that short that wouldn't be that hard to do, some camera tracking and a matte painting put together in Photoshop. The robot animation shots break down based length with camera tracking and the lighting adding more time. There are lots of numbers on how much animation an artist can make in a week or a day. But on a indie style budget you might be able to get away with 8-10 hr/work per second of final film. Hollywood style you might get a second a week (40-60 hrs). But again quality varies greatly. Robots are a bit easier to animate than organic stuff. And if you are willing to settle for what you can get with canned mo-cap/alter script to fit what is available you can get the costs down. You have to pay for the rigging and setup costs but if you use only one or two animated models that overhead only hits you for the first shot.

That final end fight with the two creatures was edited to keep the VFX costs down. The camera is always locked for each fight shot. Animating two robots fighting is a lot of work though, it's not just twice as much work as one character, the interaction work costs a lot. Add in the VFX on top (sparks, damage, etc) and you can see why they did the edit the way they did. Unless you are pretty involved with the VFX work though, that kind of short-cut may not be available. It's why you often will see someone with a VFX background doing these kinds of shorts, it's the only chance of controlling to costs. If I do something and it takes 40 hours to composite, I pay myself nothing, but if I were billing it out, you'd be looking at real money very fast.

FilmBoy77
05-15-2014, 10:39 AM
Great info Matt -- question about Turbo Squid's website. Can those 3d models (that are listed as animated) be animated in AE using the 3d Elements plugin by any chance?

msorrels
05-15-2014, 08:06 PM
Great info Matt -- question about Turbo Squid's website. Can those 3d models (that are listed as animated) be animated in AE using the 3d Elements plugin by any chance?

3D Element really doesn't have features to do the import of a rigged model. It also has very limited animation capabilities. For OBJ models bought at Turbo Squid you "might" be able to import them as prop/static objects. But I've found that 3D Element's obj importer is kind of weak. I exported a bunch of test objects (some people, some cars) and had little luck getting them to import correctly. And it doesn't read surface files at all, so you have to re-do all the surfacing entirely.

If you wanted to use models from any of the 3d marketplaces, or just free on the net, you'd need to use a bit more powerful 3d package. Which generally means Maya, 3DS Max, Lightwave, Modo, Cinema 4D, DAZ Studio, Poser, Houdini, Carrara, Blender, etc. There are more 3d packages these days than ever before. Blender could work with many of the models at Turbo Squid and it's free. Just be sure whatever model you use has a license that would allow you to do what you want. Lots of free models have non-commercial use licenses. Some paid models have exceptions on what it can be used on.