Converter creates solid black models

[pjcozzi]

Some models will generate a fragment shader that looks like

vec3 diffuseLight = vec3(0., 0., 0.);
// ...
diffuse.xyz *= diffuseLight;

which makes the model render as solid black. I suspect this is because the COLLADA model is incomplete, but the converter needs to be more robust here since many models in the wild are like this.

For example, export this NASA Blender model to COLLADA and convert to glTF: http://www.nasa.gov/multimedia/3d_resources/assets/eoss-messenger.html

Here is the converted model using dev-4: https://docs.google.com/file/d/0ByQK8VRy-TyUaTR3dFcyelpTVW8/edit

For more discussion, see: https://groups.google.com/forum/#!topic/cesium-dev/soK05NfUuM8

[pjcozzi]

This came up again with another NASA model, again using Blender.

discussion: https://groups.google.com/forum/#!topic/cesium-dev/WoiWRopEQJQ

model: http://www.nasa.gov/multimedia/3d_resources/assets/jpl-vtad-ds1.html

[fabrobinet]

looking at the model just above, what generates blender is surprising/strange. 2 lights of type ambient are created but they are heavily customized with extras to act as hemisphere lights...

[pjcozzi]

Should we write a bug Blender bug? Should we use different export options?

[fabrobinet]

@pjcozzi well yes, a bug should be filed, I guess they use ambient as a placeholder to put their custom light. But even doing the same sort of thing, they should probably use omni or directional to provide at least a light for which the transforms that are applied have some sense and hence get a closer rendering for the viewers not supporting their extra.

[fabrobinet]

Fixed in dev-6.

[pjcozzi]

OK - when do we expect to merge dev-6 to master?

[fabrobinet]

@pjcozzi I am fine doing a small release for dev-6 that just contains bug fixes (and no format changes) if you prefer to get it quickly, as I understand some issues impact cesium support. Before doing so I want to fix a couple of bugs like this one: #244 and also set properly version as discussed in #229 .
Hopefully this should happen during next week.

[pjcozzi]

Yes, I am totally for a bug fix release even without any format changes.

[fabrobinet]

OK.

[antont]

good to hear that dev-6 is indeed in good shape for master - i was just reading the logs / diff in order to know whether should use master or dev-6 for a project that begin on monday .. if indeed choose to use gltf there (with three.js).

btw about Blender Collada woes & dev there is an on-going discussion now on the mailing list. Most recently it has been maintained just by a single volunteer (Gaia) for second life export support (actually the feature set there is probably quite similar to gltf). But she's gotten tired of the responsibility and would like to drop it. Blender Institute guys are not keen on Collada and are now working on reverse engineering FBX to implement basic support it in Python, with funding from Valve for Steam modding community actually. But Mixamo has stepped in and shown interest in Collada (animation import) improvements, we'll see if that actually leads to something.

I mentioned our recent new interest & successful experiments with Blender Collada exports for glTF use. And indeed if we start using glTF for biz then we'll depend critically on that Blender export.

A couple perhaps interesting posts: 1. Blender's Campbell's views & info about FBX work http://lists.blender.org/pipermail/bf-committers/2014-April/043378.html and 2. Mixamo's defence for Collada referring to improved Autodesk support etc. http://lists.blender.org/pipermail/bf-committers/2014-April/043389.html

[pjcozzi]

Thanks for sharing this info and links. Very interesting stuff. I certainty hope COLLADA support, export in particular, will improve in Blender.