SceneObject is header file only.

ModelViewerWindow loads model as follows:

GPUModel* newModel = loadModelFromOBJFile(filename, errorMessage);

… and collada / dae file as follows:

ColladaLoader colladaLoader;
Gnz::SceneObject* scene = colladaLoader.LoadFile(filename.toLatin1().data());
scene->applyAnimations();

• add VertexDeformRenderPass to RendererFactory
• GnzQQWidget child class with extra properties for json description.

From modo we get a whole bunch of locators and animations that operate on them. All of this is loaded into the ModelViewer. Shader code uses the locator animations to warp the geometry using a vertex shader.

There's more in the Collada: geometry, lights, cameras, etc. (the whole render object tree in Modo is there)

Lattice deformer tutorial on YouTube: http://youtu.be/jviQSB6quJY

Read up about OS Mesa – llvmpipe is the new software rendering pipeline that compiles shader code using llvm.

In GnzExamples/ModelViewer/VertexDeformRenderPass I have to modify the following line so that we can see the surface:

out_Normal = -normalize(vo_Normal)

ColladaLoader loads collada DAE file into SceneObject

SceneObject holds all Animation objects and SceneNode hierarchy.

Had quick look at GnzExamples::ModelViewer* – not yet clear where the deformation is currently happening.

BlendRenderPass, TextureToScreen set viewport offset. GBuffer sets viewport but no offset.

Should pick convention: either whole pipeline does not set offset and the final stage does, or everything sets offset.

Based on:

vec4 out_FragColor = vec4(gl_FragCoord.x / 1000, gl_FragCoord.y / 1000, 0, 1);

it looks like gl_FragCoord is not relative to moving viewport, but absolute: it increases as the animation moves the viewport to the right and to the top.

ivec2 Base = ivec2( (gl_FragCoord.xy - u_ViewportXY) * StepSize);

• Slider node
• TwistDeformer node on the GPU (Vertex shader)
• SineDeformer node
• Multiply node