just a cube with 3.5 ml polygons
export glb - 746 mb, 10 minute ++
export fbx - 155 mb, ~1 minute
its not a bug, but it is impossible to work with this format
We'll look into this. In the meantime, you can significantly reduce the export size by disabling any of the following mesh data you do not need:
In addition, compression can be applied to a glTF file using tools like glTF-Pipeline, which should reduce the size of the geometry by 85-95%. For very predictable geometry, like cube/voxel stuff, compression can sometimes do better than 99%.
I would even say the size is not important, as the time for export / import
@koloved Can you please attach an example file? Even if the export time was higher than for fbx, it was still less than factor 5 and the glb was actually smaller than the fbx in my test.
@UX3D-schmithuesen
gltf - 1 hour and more! =)
fbx - less 5 seconds
the file (apply modif, select only) - file
_it is completely unusable_
@koloved This was just fixed today in #158. It will probably be a day or two before you see the fix appear in the next Blender build.
With the latest fixes, and "apply modifiers" checked, the export here now takes about 45 seconds on my machine. I confirmed that FBX (with the modifier manually applied) takes less than 5, so, there's still room for performance improvements here, but at least the hour-long hang is fixed.
File size is also just double for me so I wouldn't say that things are problematic anymore.
personnaly i will not use glTF while it's not fully implemented into Maya
( i mean with a working exporter and with an importer )
@anto2102 the most effective place to comment on this issue would probably be https://forums.autodesk.com/t5/maya-ideas/gltf-support/idi-p/7958727.
Just wanted to leave a note -- maybe there's a regression, but on Blender 2.82 on Ubuntu, importing a 50mb FBX and rexporting out to GLTF gets us a 500 MB file. The file has lots of blendshapes, though (it's a face), maybe it is something to do with blendshape data specifically, because I also just kicked out some static scene GLTFs with no issues.
Weirdly had the same experience on another dataset where we converted an OBJ to GLTF using the Cesium tool, and it went from 10 mb to 100 mb, which should literally never happen. Wonder if this is a GLTF lib issue, not a blender issue.
Please share example files — there are many reasons a file’s size may change.
@donmccurdy
Seems to be related to Blender, not GLTF
Just wanted to leave a note -- maybe there's a regression, but on Blender 2.82 on Ubuntu, importing a 50mb FBX and rexporting out to GLTF gets us a 500 MB file. The file has lots of blendshapes, though (it's a face), maybe it is something to do with blendshape data specifically, because I also just kicked out some static scene GLTFs with no issues.
Weirdly had the same experience on another dataset where we converted an OBJ to GLTF using the Cesium tool, and it went from 10 mb to 100 mb, which should literally never happen. Wonder if this is a GLTF lib issue, not a blender issue.
I am having a similar issue. I have a 17 MB .glb which i import into blender to adjust and change the origin. On export the same model now has a file size of 90+ MB. It would be great if you (@shawticus ) could give any pointers regarding what settings to look for in Blender.
@v-prgmr if this is still happening on a recent version of Blender, could you file an issue with an example .glb attached? Thanks! We cannot investigate these issues without examples to reproduce the problem.
@v-prgmr if this is still happening on a recent version of Blender, could you file an issue with an example
.glbattached? Thanks! We cannot investigate these issues without examples to reproduce the problem.
@donmccurdy Will do so 👍
Thanks for the quick response
(for future readers: https://github.com/KhronosGroup/glTF-Blender-IO/issues/1166)
Most helpful comment
@koloved This was just fixed today in #158. It will probably be a day or two before you see the fix appear in the next Blender build.
With the latest fixes, and "apply modifiers" checked, the export here now takes about 45 seconds on my machine. I confirmed that FBX (with the modifier manually applied) takes less than 5, so, there's still room for performance improvements here, but at least the hour-long hang is fixed.