We currently generate uniforms for custom parameters when loading .brdf files, but right now there is no way for the user to control them. We should instantiate sliders in order to do this.

See the black circle around the +z axis (blue line).

Some shaders, like ashikhman_shirley appear very dark with just a point light. Might be nice to be able to change light intensity.

Right now the checkboxes for analytical BRDF parameters look out of place. They are especially small in Chrome. They should look better and perhaps be bigger / easier to hit on mobile.

Wondering if we could rotate objects around their centroids instead of the world origin. In the picture above, objects rotate out of view.

The GUI should not need to be aware of the Model. Furthermore, Viewports should not need to be aware of the GUI. In other words, there should be a "one-way" data flow from GUI to Model to Viewports. @n8vm tells me that this is sort of like the "MVVM" paradigm.

Firefox throws the following warning on launch (on master):

Error: WebGL warning: drawElements: This operation requires zeroing texture data. This is slow. ModelViewport.js:578:7

There should be a way to click on the teapot and superimpose the BRDF lobe.

On the lobeRdr_bug branch, see commit e5387a4015391cef7ba02d61294923684b95a16b.

If we click and drag across the teapot, lobe glitches out. However, if
we uncomment line 988 of ModelViewport.js, then it works.

More specifically, calling gl.bindBuffer on line 253 of LobeRenderer.js causes the breakage.

I'm not sure why this is happening because both lobes are in completely
separate OpenGL contexts and therefore one does not write to the data or
state of the other.

They still aren't separated yet.

As of now it only seems to reset when moving the camera.

Right now when our slider values transition between single-digit and double-digit values, there is no formatted print so it changes the location of the next slider, which causes jittering. We should be able to fix it using string formatting, or by putting sliders in separate rows.

Do we want a noise-free visualizer? (Might look good in a side-by-side comparison against Disney's tool.) Do we want to showcase realtime techniques? If we want to do those things we should investigate SH.

References:

1. Stupid Spherical Harmonics Tricks, PP Sloan, 2008
2. Unreal Engine SIGGRAPH talk, 2013
3. An Efficient Representation for Irradiance Environment Maps, Ramamoorthi and Hanrahan, 2001

The current lobe visualizer assigns a color based on normal direction. We should move to Phong shading. Also, we could make it look cool by using an intensity-based colormap, where longer directions have a different color. (This is what BRDFLab does, see screenshot below).

We should be able to choose .brdfs on the server rather than forcing the user to upload their own.

We will need to turn PointlightViewer's fragment shader into a template, and then call the inlining code to substitute the analytical BRDF into it. (We are already doing the same thing in BRDFViewer's vertex shader.)

Workaround: set dom.moduleScripts.enabled to true in about:config.

For some reason, with IBL, the radiance map does not appear on Android (it just draw black).

TODO: Figure out how to get debug messages out of Chrome on Android.

It would be nice if we had some more BRDFs to play with. It should be easy to add more BRDFs in Disney's format - they just have to be modified slightly to conform to OpenGL ES rather than desktop OpenGL. The easiest way to start is to load a Disney .brdf file, see what breaks, then change it.

Disney's files are here: https://github.com/wdas/brdf/tree/master/src/brdfs .

Perhaps this task could be automated with a script? I believe that in the general case, converting desktop OpenGL to OpenGL ES is nontrivial. Also, the total number of files that Disney has is small, so it may be faster to just do it by hand than it is to write a foolproof script.

Seems to appear near phi=0. Does not appear to be a Mach band. TODO: Add a screenshot.

Load the model from a GLTF file, and also load PBR properties (if any) and visualize the BRDF.

Ideally, we should be using the function form of "use strict". Currently I'm overriding JSHint to ignore this.

Some variables and function names use underscores, other use camel case. We should be consistent. This is pretty low priority.