convenient handling of complex 3d scenes (DX11 & EX9)
Ever got tired of binsized SetSlice, Sift, * (Transform) mayhem when using the collada/assimp nodes?
This contribution aims to solve these issues.
Collada and Assimp are both great at doing what they were intent to: loading 3d scenes. They just don't natively provide a vvvv-friendly way to actually make use of all the info they load. The mere spread outputs of transforms, meshes, textures, etc. obfuscate the hierachical relation, and forces you to juggle binsizes while keeping the differnt types in sync.
so decode.io and I put our heads together, thinking about a patcher-friendly way and came up with this.
load the scene, apply some transforms and one 'Split' like node which outputs everything nicely for a shader.
2. Convenience Node displaying a browsable SceneGraph tree with all the relational info in it's place.
3. Apply transforms by specifying which part of the graph/tree should be affected. Accumulation and propagation to all children is automagically handled in the back.
4. Model outputs all the drawing relevant data of the (sub)graph in nice little spreads to use connect directly to a shader without further ado. Each meshpart has its corresponding transfrom and material in place. For not textured parts a default white texture is inserted to avoid bin sizing. Check the config pins to configure the node to do just what you need (e.g. multiple textures, bouding box, bone tranforms...)
The Scene (=loader node) outputs the type GraphNode which is a reference copy of the whole scene hierarchy, namely SceneGraph. Transformation can be applied on any node (or group or spread of nodes) within that graph.
in case you can draw everything at once with the same shader. you're done at this point.
If not the Selector lets you move a 'pointer' in a GraphNode another descendant than the root, thereby you select just parts of the whole model for drawing and further transformation downstream.
(more implementation details on demand)
supported by decode.io
possible further development features
Superb! Thanks WOEI and DECODE!
This is huge and provides a solution to a problem that's been a pain in vvvv, while environments like Unity offer way smoother and integrated asset handling/scene explorers.
Exactly what I had in mind when writing this:
Still, it doesn't solve the problem that these approaches don't ship with vvvv/don't integrate well as there's no proper interface to load those modules up but to manually place them in your patch.
Anyway, I very much welcome this. Finally something that aims to provide high level user friendliness.
Just a couple of vvvv nuisances left in this regard. I will address this at the summer camp.
noice! this should be pull requested into the DX11 pack
that saves some tedious patch work, cheers!
Awesome, thanks! glTF support with PBR parameters would be great! any way to support this?
Looks very useful :)
@mburk: sorry, completely missed your comment! if you're still interested, let's talk
Hi , sorry if this is a stupid question , but as there is no "install" section , where do i extract this zip ? my vvvv folder is a mess and i cant get the node to popup when i type its name (i assume it s not in the "packs folder" then ?)
hi, sorry. forgot to add install instructions. it's a regular pack, so installation is the same as for dx11 et al. just unzip inside vvvv then windows will integrate the content into the existing packs folder.
.../vvvv_50betaxx/packs/struct/... is how it should look like in the end.
Hello, I am not able to use the node animate correctly. None of my 3d models that work with Skeletalmesh (dx11 assimp) are being animated. What am I missing out? thank you
@felipelinsf create a forum post and attach one of the models that doesn't work.
I can not attach files in the forum because I am a new member, but for example I can not animate the model astroBoy_walk.dae that comes in the vvvv installation folder
SceneGraph animation nodes are meant for plug and play with mesh transforms. skeletal animation is possible, not with plug and play support tough, but with some more patching required:
the Model node has hidden pins to give you bone information, this is where the skeletal animation info is. you need to animate those, combine them with the local transformation of the mesh and feed it into skinning matrices which need to go into a shader which supports them.
Do you know of any examples where bone matrices are animated in vvvv? It'd be nice to have full real-time control of the model. Thanks
When I try to open any simplecar.fbx example, it doesnt work
00:08:15 : cannot load C:\vvvv\vvvv_beta_39_x64\packs\SceneGraph\assets\simplecar.fbx
00:08:15 ERR : System.ArgumentException in System.Xml.Linq: '' is an invalid expanded name.
at System.Xml.Linq.XName.Get(String expandedName) at SceneGraph.Core.GraphTraversing.ToXElement(GraphNode node) at SceneGraph.Core.GraphTraversing.ToXElement(GraphNode node) at SceneGraph.Core.Scene`1.CreateGraph() at SceneGraph.Adaptors.SceneAssimp.CreateGraph() at VVVV.SceneGraph.LoaderSceneGraphNode.Evaluate(Int32 spreadMax)
the guys at decode.io did a proof of concept patch. i'm sure they'll pass it on if you ask them
can't reproduce. did you alter the fbx file? because the error sounds like it is loading an object within the scene without name.
Danged if I can find a way to get the texture transforms. If I load a model that has a repeating texture applied, the texture is stretched to the whole mesh since there is no way to apply anything to the Texture Transform pin of a shader.
The Assimp nodes suffer from the same problem - there is no Texture Transform output pin on the Materials node there either. So is this just an artifact of the underlying library not supplying it?
I was all excited to use this since it is so much more elegant than the native Assimp nodes, but this is a deal killer. Pout. Am I missing something?
anonymous user login