Model-viewer: Maps moiré effect

Created on 18 Dec 2019  Â·  13Comments  Â·  Source: google/model-viewer

Description

Our objects have a moiré effect on certain maps. It is the same with glb and gltf.

Is there a solution for this?

Live Demo

https://framework.tv/leistungen/digital/webar-obi-uebergabedaten/

Browser Affected

  • [ X] Chrome
  • [ ] Edge
  • [ X] Firefox
  • [ ] Helios
  • [ ] IE
  • [ X] Safari

OS

<

  • [ ] Android
  • [ X] iOS
  • [ ] Linux
  • [ ] MacOS
  • [ ] Windows

Most helpful comment

@Sbstnlhmchr Thank you for the example! The problem is with the model, but it's very easy to fix (and oddly common; may I ask what program created the glTF?). The samplers are set to LINEAR interpolation: "minFilter":9729, but you want LINEAR_MIPMAP_LINEAR: "minFilter":9987. With that change to both samplers, the Moire is gone.

All 13 comments

Looks like incorrectly generated mipmaps to me - also curious about this.

I believe I fixed that in #880, which is slated to be released imminently in 0.8 (we're just finishing up a few bug fixes at the moment). Can you try master and see if it's improved?

@elalish I took the well-known route to installation because I don't know exactly how to use master. We got the problem under control by deleting all maps (normal, metallic, AO) and reducing the contrast of the diffuse map by approx. 70%.
At usdz we kept all maps.

Thanks for your help

@Sbstnlhmchr Oh my, please don't delete all your maps! We just released 0.8 yesterday, so that should make it much easier to test. If any rendering problems remain, please let me know!

@Sbstnlhmchr can you confirm that you are (or are not) still experiencing this issue in our latest release (v0.8.1)

@cdata hello, thanks for asking. I turned off the url because the project is submitted. To solve the problem, we reduced the contrast and deleted almost all maps. It is not yet optimal, but it looks good - the URL is online again for two days.

No problem, hopefully we can help discover the problem. Would it be possible for you to share a screenshot (or better: a model) w/ maps?

https://framework.tv/wp-content/uploads/2020/01/Problem-Moire.zip
Here is an example created from C4D. It is just a cube and has a NormalMap and DiffuseMap.

@Sbstnlhmchr Thank you for the example! The problem is with the model, but it's very easy to fix (and oddly common; may I ask what program created the glTF?). The samplers are set to LINEAR interpolation: "minFilter":9729, but you want LINEAR_MIPMAP_LINEAR: "minFilter":9987. With that change to both samplers, the Moire is gone.

@elalish we are also having this issue. It wasn't apparent to me right away but it amplifies when using multiple models on a single page.

our models are using the obj2gltf tool by cesium.

I filed a bug report there:
https://github.com/CesiumGS/obj2gltf/issues/239

@elalish got a reply from obj2gltf developer:

image

would that work?

Yes, removing the default would be ideal. Technically that's not letting
the engine decide, as the glTF spec specifies which default the engines
must use. However, they chose that default with input from the engines, as
we know anything other than LINEAR_MIPMAP_LINEAR results in significant
rendering artifacts. The performance difference from this is small since it
is handled in hardware anyway.

Thanks,

-Emmett

On Sun, May 3, 2020 at 6:41 PM Markus Lanxinger notifications@github.com
wrote:

@elalish https://github.com/elalish got a reply from obj2gltf developer:

[image: image]
https://user-images.githubusercontent.com/20549552/80931297-70f29b00-8deb-11ea-90c9-62110cd65bc7.png

would that work?

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
https://github.com/google/model-viewer/issues/936#issuecomment-623223376,
or unsubscribe
https://github.com/notifications/unsubscribe-auth/AAMS2LDGUWICY2L6V5CLPV3RPYMN7ANCNFSM4J4IGDLA
.

thank you @elalish

Was this page helpful?
0 / 5 - 0 ratings