3D netball outfits

[ozRocker]

I just finished off a website for a company that creates sports uniforms.  They wanted something specifically for netball uniforms.  They originally had a 2D dress designer but I convinced them to go for 3D

http://dyo.100netball.com.au/

 

[jerome]

excellent !   as usual ;-)

[ozRocker]

I asked the client to send one of the netball dresses so I can use it as the base dress.  I don't care about texture, only the geometry is important.  So I said "Send me your cheapest one that represents the shape".  I took 82 photos of the model in that outfit in an A-pose.  Girls always have very clear skin, not enough features, so I put dots on them using eye-liner.  If the clothing lacks features I add masking tape with crosses on them.

[ozRocker]

I had to clearly separate dress from body using the knife-tool in Blender.  Then make separate materials for body and dress.  Client already had the outfit designs in PNG so I had to repologise dress and create a UV map that would fit the existing outfit designs.

I used a normal map for the dress only to save memory.

I just used standard animations from Mixamo for idle and shuffling.

[GameMonetize]

Impressive!

[JCPalmer]

Really good.  I deal primarily in makehuman meshes, not got the stuff or expertise to do that.  The breathing was something that I wish I could do.  I probably could if I messed with the bones, I guess.

Going from still photos firmly shows that this is not a "cartoon".  The non-blinking blank stare is a trade-off, but I am probably the only one that would notice. 

Since you are moving the lighting with the camera (I remember that thread & do it too), I  think the shadows from the same light are counter productive.  Having a second stationary light with intensity = 0 would make your light source appear to be coming from the same point at all times.  The zero intensity means it will have no effect, except to the shadows.

[ozRocker]

14 minutes ago, JCPalmer said:

Really good.  I deal primarily in makehuman meshes, not got the stuff or expertise to do that.  The breathing was something that I wish I could do.  I probably could if I messed with the bones, I guess.

Going from still photos firmly shows that this is not a "cartoon".  The non-blinking blank stare is a trade-off, but I am probably the only one that would notice. 

Since you are moving the lighting with the camera (I remember that thread & do it too), I  think the shadows from the same light are counter productive.  Having a second stationary light with intensity = 0 would make your light source appear to be coming from the same point at all times.  The zero intensity means it will have no effect, except to the shadows.

Makehuman is awesome!  For perfect retopology I'll use Makehuman meshes as the base mesh when I use Wrap (http://www.russian3dscanner.com/)  They have great UV maps too.

The breathing is just an "idle" animation from Mixamo.  The non-blinking stare annoys me more and more when I look at it.  I've had some practise with morph targets with Babylon.js so I'm going to experiment with facial expressions to see if it doesn't chew up too much memory or blow out filesize.  I still need all this to work on mobile devices.

 

[JCPalmer]

I have been able to get expressions for makehuman to even expressively talk,  see topic.  I use my own morph targets for the QueuedInterpolation extension I am doing.  This does mixing and animating of targets on the cpu.  The morph targets can be much smaller due to being on the cpu, allowing sets of targets be on only a subset of verts, called shapekey groups.  You may also have multiple independent groups on the same mesh, e.g. FACE, EYES, EYELIDS, LEFT_HAND, RIGHT_HAND, BREASTS.  I store 28 very compact keys that are manipulated by those sliders, and 4 eye buttons.  My keys can have both a positive & negative value, and also be greater than 1, or less than -1.  The negative values are defined against a single axis which you pick, so you can really make many consolidated keys from fewer file loaded keys.

BJS's implementation is GPU based, and can mix keys.  Each key contains the verts of the entire mesh, and are not in as compact a representation on disk.  Think the retopology you speak of makes the process from makehuman I use very difficult though.

[ozRocker]

YAY!  This made the front page of Babylon.js!!!   Thank you so much whoever put that there.  It is a great honour

[GameMonetize]

We are the one who are honored!