jem

I cannot explain that error. It looks like the UV map for the faceplate is out of sync with your AO map. I assume you just fixed your UV map and AO bake for the last issue, so I cannot see why they would be out of sync.

The black area in the AO map near the lock should be completely obscured by the lock body if the alignment was correct. The point being, there shouldn’t be much AO visible around the lock because of the obtuse angles involved between the bodies. In a pinch, you could paint out the lock AO on the AO map and figure out the real cause later. Sorry, I know that is a cheat.

Thank you, Verge3D team. These enhancements are great! I had not even considered using puzzles to create geometry from scratch.
Also, the “on error” callback slot on the loader puzzles will make much more robust solutions.
Thanks again.

@gurlaldeep singh, if you are ever in doubt about which files are used in the browser by a Verge3D scene, pop open Chrome dev tools and look at the network tab. Dev tools will show you which files are used by the scene.
If you are like me and have browser extensions installed, there will be a lot of noise in the network tab. Sort the list by domain, and your files will be grouped.

@kogoshi, as long as the Maya version has a value node and math nodes, you can do this. The math is easy. I will assume that Maya uses a 0…1 scale for values.
1. Get a value from the puzzles. The value must be 1 or 0. O indicates no-change. 1 indicates inversion.
2. Subtract the diffuse value from the inversion value.
3. Take the absolute value.

Here is the node set up in Blender.

@visualizer, your work is terrific! Making convincing-looking glass is difficult. I am not an expert but here are a few thoughts.

The scooter’s windscreen is curved, and your watch crystal is flat. The curvature difference will cause the reflections of the HDRI to play differently on these two objects. Reflections are more interesting on curved surfaces. The edges of your watch crystal are sharp. Would it be inappropriate to give the edges of the crystal a small bevel and shade it smooth? If you could, the crystal would pick up many more glints of light from the HDRI.

Keep in mind that glass is never 100% clear; even a sapphire watch crystal. If you look at the scooter glass, you will see that the Glossy BSDF node has an RGB hex value of E7E7E7, a very light grey. This gives the glass a little color and presence even when it is not reflecting an HDRI light. I think that you are doing this. When I look at the watch crystal from the side, I can see the fresnel mixing in a darker color.

The other concept that stumbled across is that glass is never 100% clean. Even if I am building a configurator for a new product, I have found that mixing in the slightest grunge map into the glass shader adds realism. But, in my case, I am configuring machinery and not jewelry. The glass is not the focus of my 3D scenes. In your case, this technique may not be acceptable.

Lastly, you can try different HDRIs. The stock HDRI is very dramatic, but it does not have a lot of detail. Unless the watch crystal is pointed at one of the white lights in the HDRI, there is not much of a reflection.

@web, yes, I can confirm the behaviour that you have described. I did try this in Firefox for Android also. On FF, the blur behaviour is the same as on Chrome, but the browser bar issue is not present. I suspect that the black bar issue is a bug in Chrome for Android.

Those are interesting solutions, but I think that this community focuses on dynamic 3D content. As soon as we export the 3D content from Sketchfab, Adobe, or some other exporter to USDZ, we might as well have chiseled the 3D object from stone. We lose the ability to drive the shape, texture, and animations of our 3D objects from puzzles or user selections in our web pages. USDZ might be a good solution for a post-configuration download file for iOS users, but it is not useful for interactive content.
Again, I am going to blame Apple for eschewing standards and trying to wall off their customers from the wider, more interesting world (and I am one of their customers).

@web, the WebGL 1.0 spec says that both the length and the width must be powers of two, so 2^x by 2^y is okay.

Non-powers of two textures (NPOTs) are supported by WebGL 2.0, but Apple does not support them.

IMHO, you should scale the images yourself before you post them. This gives you control over how the final image looks. If you don’t scale the images yourself, either the Verge3D engine or the video driver will do it and the outcome is not under your control.