IGA Hair Salon : How we do it.

  The IGA campain features anywhere from 3 to 16 characters per spot.  All these CG actors need to drop by the virtual hair salon before they are allowed on set.  Here's what happened to Oceane Rabais and Bella Marinada at this stage.

1-We always start with the character design made here at SHED as a reference.

  For the whole hair process we use a collection of in-house compounds that are derived from krinstinka ( http://www.matkovic.com/anto/kristinka-hair.html ) and Melena (http://opensource.nestanimation.com/melena.html ).  We also recently added a few nodes fromTriggerfish animation studios ( http://www.triggerfishstudios.com/en/ ).

2 - We then look up on the internet for a real life reference of what the hairdo could look like.  This is only as a reference to capture certain real life details.  Since we are going for a Cartoonish look, we are not aiming at reproducing the reference exactly.  Of course a picture of a duckface girl is always a plus.

3 - We proceed to create an emitter fitted to the head from which we emit guide strands with Ice.  They get their shape from nurbs surfaces.  Those guides are low in number (from 200 to 400), so it's easy to work with them to groom and later simulate and cache on disk.  The idea is to get the shape of the hairstyle and the length.  The bright colors are there to help see what's going on.

4 - Next, we clone theses strands, add an offset to their position and apply a few Ice nodes to further the styling.  These nodes generally include randomizing and clumping amongst others.  We now have around 90 000 strands and it can go up to 200 000.

5 - Then we repeat the process with the eyelashes and the eyebrows.  During the whole process the look is  tweaked in a fast rendering scene. 

6 -  Once happy with the results, we copy the point clouds and emitters to the "render model" where the point clouds will be awaiting an Icecache for the corresponding shot.  We use Alembic to transfer animation from rig to render model and the Ice emitters are "cage deformed" to the alembic geometries because the hair styling is done too late in the process to include theses emitter in the alembic export.

7 - Back to the Hair model we convert the guides strands to mesh geometries.  We apply syflex cloth simulation operators to these geometries to get ready for shot simulation.  We link the guide strands to the syflex mesh so they inherit the simulation.

8 - Next comes shot by shot simulation and Ice caching of the guides strands (hair, lashes, eyebrows and beard if necessary).

9 - Before we pass down the simulation caches to the rendering department, we need to do a test render to be sure every frame works and there is no glitch/pop.  With final beauty renderings taking sometimes close to 2 hours per frame, it is not a good thing to have to re-render a shot because a hair strand is out of place !  The scene we use renders quickly with no complex shaders and only direct lighting.

10 - Once we are happy with the look of the hair, the movement of the simulation AND most of all once we've resolved all the problems, we give the signal to the rendering department.  The hair PointClouds are always automatically linked to the appropriate simulation cache for the current shot so all they have to do is "unhide" the corresponding object in their scene and voila !

Luc Girard, our hair artist, was interviewed by the guys at TD Survival on facebook. You can watch the video here:

Workflow Tips - Hair & Grooming from TD Survival on Vimeo.