[x3d-public] got sample X3D model for FontLibrary or Tangent? glTF conversion progress, tooltips, names, etc.
Holger Seelig
holger.seelig at yahoo.de
Sun Jan 11 12:27:55 PST 2026
Here is a very simple example with only one shape that uses MultiTextureTransform (two different TextureTransformMatrix3D nodes) and MultiTextureCoordinate (where both TextureCoordinate node are actually the same). I have added a SheenCloth.x3d, converted with X_ITE. I have modified the original conversion file to make it work in Castle Model Viewer (removed the SheenMaterialExtension node, added proper NavigationInfo and Viewpoint, added MultiTextureCoordinate because Castle needs this, X_ITE uses the last TextureCoordinate if a mapping does not match). I attached the whole folder with all textures and the converted file SheenCloth.x3d.
This file is originally from the glTF-Sample-Assets from Khronos:https://github.com/KhronosGroup/glTF-Sample-Assets/tree/main/Models/SheenCloth
Best regards,Holger
Am Sonntag, 11. Januar 2026 um 04:45:19 MEZ hat Don Brutzman <don.brutzman at gmail.com> Folgendes geschrieben:
[cc: x3d-public since this topic has ballooned to be much bigger than originally expected]
Background: working on putting two of Michalis' Castle Model Viewer examples for X3D 4.1 Tangent node into the X3D Example Archives.
- X3D Example Archives: X3D4AM, X3D for Advanced Modeling, Gltf Sample Models, Cat By Muru
- https://www.web3d.org/x3d/content/examples/X3dForAdvancedModeling/GltfSampleModels/CatByMuruIndex.html
- https://www.web3d.org/x3d/content/examples/X3dForAdvancedModeling/GltfSampleModels/CatByMuruX_ITE.html
and
- X3D Example Archives: X3D4AM, X3D for Advanced Modeling, Gltf Sample Models, Halloween Pumpkin Lantern Knight
- https://www.web3d.org/x3d/content/examples/X3dForAdvancedModeling/GltfSampleModels/HalloweenPumpkinLanternKnightIndex.html
- https://www.web3d.org/x3d/content/examples/X3dForAdvancedModeling/GltfSampleModels/HalloweenPumpkinLanternKnightX_ITE.html
=================
Continuing, mostly finishing this long thread:
Thanks Holger for the TextureTransform technique. I have applied that change in two models (CatByMuru.x3d and HalloweenPumpkinLanternNight.x3d to good effect.
Have also added the following tooltip.
- X3D Tooltips: TextureTransform
- https://www.web3d.org/x3d/content/X3dTooltips.html#TextureTransform
Hint: image flip horizontal <TextureTransform DEF='FlipHorizontal' scale='-1 1' translation='-1 0'/>Hint: image flip vertical <TextureTransform DEF='FlipVertical' scale='1 -1' translation='0 -1'/>
Noticed that when you apply such a TextureTransform, it is a child of the parent Appearance node, and not part of each ImageTexture (like the CGE attribute). When there is an associated PhysicalMaterial and no mapping field is defined, that means the TextureTransform applies to all of the ImageTexture fields (baseTexture, emissiveTexture, metallicRoughnessTexture, normalTexture, occlusionTexture). Similarly for UnlitMaterial.
Michalis, should I leave mapping="" blank throughout, or else apply a value there (perhaps mapping='flipVertical' throughout)?
We don't appear to mention this case explicitly in the specification, but I think the prose is OK and does not allow any other interpretation. (We can add more words for clarity if anyone wants to.)
- X3D Architecture version 4.1 draft, clause 12 Shape component12.2.2 Appearance characteristics
- https://www.web3d.org/specifications/X3Dv4Draft/ISO-IEC19775-1v4.1-CD/Part01/components/shape.html#AppearanceCharacteristics
Also thanks Michalis, I have added generateMipMaps='true' to TextureProperties nodes, and handled DEF/USE.
- <TextureProperties DEF='AveragePixel' generateMipMaps='true' magnificationFilter='AVG_PIXEL' minificationFilter='AVG_PIXEL_AVG_MIPMAP'/>
HalloweenPumpkinLanternKnight.x3d model is too large for X3DJSAIL to load, unfortunately (currently 2.6MB) and so I was not able to reduce the extra-long floats in there (no more than 7 digits of precision are needed).
Michalis, the character
| dollar; | $ | U+0024 |
is not within the range allowed for XML NMTOKEN (name token) data type, and so it is not recommended for X3D model interoperability across all file encodings and programming-language bindings. See
- XML Recommendation, v1.1 (Second edition), World Wide Web Consortium (W3C), 2.3 Common Syntactic Constructs, Names and Tokens
- https://www.w3.org/TR/xml11/#sec-common-syn
DEF/USE, proto and other names in X3D Example Archive models are always conservative for best interoperability, additional suggestions always welcome. Further details are maintained at
- X3D Scene Authoring Hint, NamingConventions
- https://www.web3d.org/x3d/content/examples/X3dSceneAuthoringHints.html#NamingConventions
Giving a node DEF='head' is legal, but it is confusing for the XML encoding, so I usually change things like that for clarity... which is often essential in the X3D Example Archives. Not sure if we should add 'head' to the list of reserved words.
Both the CatByMuru and HalloweenPumpkinLanternKnight models are looking good and looking consistent, both for Castle Model Viewer and X_ITE/Playground/Sunrize. Bravo!
X3DOM looks quite different (perhaps differing application of PBR textures in Physical Material node). Perhaps someone wants to work on that open-source codebase... examples in the X3dForAdvancedModeling GltfSampleModels directly provide a lot of excellent test cases for glTF compatibility, with many others working already.
New issue: since multiple TextureTransform nodes might be needed within an Appearance, should we allow multiple TextureTransform nodes in Appearance? Right now that is an SFNode field. If you think there is a use case that an author (or converter) would want more than one TextureTransform in the Appearance node, please advise.
- 12.4.2 Appearance
- https://www.web3d.org/specifications/X3Dv4Draft/ISO-IEC19775-1v4.1-CD/Part01/components/shape.html#Appearance
Appearance : X3DAppearanceNode {
SFNode [in,out] acousticProperties NULL [AcousticProperties]
SFFloat [in,out] alphaCutoff 0.5 [0,1]
SFString [in,out] alphaMode "AUTO" ["AUTO", "OPAQUE", "MASK", "BLEND"]
SFNode [in,out] backMaterial NULL [X3DOneSidedMaterialNode]
SFNode [in,out] fillProperties NULL [FillProperties]
SFNode [in,out] lineProperties NULL [LineProperties]
SFNode [in,out] material NULL [X3DMaterialNode]
SFNode [in,out] metadata NULL [X3DMetadataObject]
SFNode [in,out] pointProperties NULL [PointProperties]
MFNode [in,out] shaders [] [X3DShaderNode]
SFNode [in,out] texture NULL [X3DTextureNode]
SFNode [in,out] textureTransform NULL [X3DTextureTransformNode]
}
Have fun using glTF models in X3D! 😀
all the best, Don-- X3D Graphics, Maritime Robotics, Distributed SimulationRelative Motion Consulting https://RelativeMotion.info
On Sun, Jan 4, 2026 at 1:11 PM Holger Seelig <holger.seelig at yahoo.de> wrote:
Instead of a flipVertically field, it is also possible to achieve this with a TextureTransform node
DEF FlipVertically TextureTransform { translation 0 -1 scale 1 -1}
Another solution is to flip all texture coordinates (y’ = 1 - y).
After all investigations I think that Castle and X_ITE are on the right way representing glTF textures, but all the examples miss a flipVertically in either way.
Best regards,Holger
—Holger Seelig
holger.seelig at yahoo.de
Am 04.01.2026 um 01:32 schrieb Michalis Kamburelis <michalis.kambi at gmail.com>:
Hello Don,
Great, thanks for the analysis of the file and tests on other browsers! Answers:
1. flipVertically:
- This is documented on https://castle-engine.io/x3d_implementation_texturing_extensions.php#section_flip_vertically , with reasons why.
- We talked about it in the past on x3d-public already.
- No, I don't think now that it deserves to be added to spec. Let it remain CGE extension. CGE should export with TextureTransform making this "flipVertically" unnecessary, and the files spec-conforming. Implementing this by "flipVertically" in CGE is just an optimization (see link above), effectively, since it's zero overhead to flip at loading. I feel it can remain CGE extension, actually.
- So, give me time, I'll fix these models to not use flipVertically :)
2. $ in name:
- I was under impression that $ is actually OK inside name in X3D. We use it in special situations (when "encoding" names with arbitrary chars from other formats) deliberately.
- Spec https://www.web3d.org/specifications/X3Dv4Draft/ISO-IEC19776-2v4.0-WD1/Part02/grammar.html about "IdRestChars" implies $ is OK. ("$" is 0x24), from what I can tell.
- CGE code derived from this is https://github.com/castle-engine/castle-engine/blob/7dccdbb455d544e3659d6f049cde9bcf428c3c86/src/scene/load/x3dloadinternalutils.pas#L82 .
- So: Are you sure "$" is not valid in X3D name? Where is this mentioned in spec? If that's the case -> no problem, I'll fix our code, and the test models.
3. """Wondering why you have those EXPORT statements"""
- Our glTF->X3D converter EXPORTs some named nodes, to allow animation of them by the outer X3D file.
- In short, it doesn't matter for these particular examples, these EXPORTS are not used in this case.
4. """<Transform DEF='head'/> has DEF name that illegally overrides a reserved word from the X3D Specification"""
- I was not aware that such names are disallowed.
- "head" is not mentioned as keyword on https://www.web3d.org/specifications/X3Dv4Draft/ISO-IEC19776-2v4.0-WD1/Part02/grammar.html .
- It's not used by classic encoding, it's only an element name in X3D XML. I don't see how node name "head" could be really confused with element <head> in X3D?
- That said, we could fix the model (and CGE converter). Can you show me where in the specification is the "head" for node name disallowed? I can add a logic to our converter to avoid "head" name then.
5. """<TextureProperties DEF=''/> requires generateMipMaps='true' since minificationFilter='AVG_PIXEL_AVG_MIPMAP""""
- I'll fix it, indeed spec says "Mipmaps are required for filtering modes with MIPMAP in their value."
- Note that CGE ignores this field. Only the filtering determines whether we need mipmaps... so to us, generateMipMaps field is actually not useful.
- (And it could be confusing in case of files like DDS or KTX, where mipmaps are part of the image. In such case, we have mipmaps, we use them -> but we don't "generate" them).
- Anyhow, that's all minor notes :) We can just place generateMipMaps='true'.
6. About the float precision: We have a --float-precision option to Castle Model Converter, it was just not used for this. I can indeed reexxport this with smaller precision.
Summary: Give me time :) and I'll fix
- AD 1 - flipVertically- AD 5- generateMipMaps- AD 6 - float precision
in the models.
For 2 other questions, I'll appreciate a link to spec :), but of course we could also adjust to them, nothing problematic there :) This applies to:
- AD 2 - $ in name- AD 4 - "head" in name
Regards,Michalis
sob., 3 sty 2026 o 07:11 Don Brutzman <don.brutzman at gmail.com> napisał(a):
Thank you very much for recovering those, Michalis!
I copied all three and put originals in version control, and began work on first. Next step, ran/committed .x3d version through X3D Canonicalization (C14N) so that any changes which occur are evident. Also added a variety of meta tags.
Validation diagnostics quickly found the following errors:
- Attribute "flipVertically" must be declared for element type "ImageTexture". (now omitted, may look funny)
- Attribute value "CastleEncoded_Pose_scene$46$fbx" of type IDREF must be an NCName when namespaces are enabled. (changed $ character to underscore _ character)
If you would like to add flipVertical and flipHorizontal attributes to ImageTexture, or to interface X3DSingleTextureNode (ImageTexture, MovieTexture, PixelTexture), please propose that on x3d-public mail list. Seems plausible to me, if glTF has something like that then let's do it consistently. No need to add "ly" suffix to names of such boolean fields. If everyone reaches consensus (or interest at least) then I can add them to XML DOCTYPE and Schema for draft version 4.1.
Continued work on other diagnostics...
Wondering why you have those EXPORT statements?
Got illegal DEF name warning, renamed 'head' to 'modelHead'
- <Transform DEF='head'/> has DEF name that illegally overrides a reserved word from the X3D Specification [/X3D/Scene/Transform/Transform/Transform/Transform[1], error]
Got several X3D Schematron warnings like this, do you agree?
- <TextureProperties DEF=''/> requires generateMipMaps='true' since minificationFilter='AVG_PIXEL_AVG_MIPMAP' mode indicates MIPMAP rendering [/X3D/Scene/Transform/Transform/Transform/Transform[1]/Transform/Group/Shape/Appearance/PhysicalMaterial/ImageTexture[4]/TextureProperties, error]
- looks like those three nodes are identical might be single DEF with one USE
Got a Schematron warning to take advantage of DEF/USE, and so did that.
- <ImageTexture DEF=''/> url array address(es) duplicate the url definition found in a preceding node, consider DEF/USE to reduce download delays and memory requirements for url content (url='"textures/lambert2_metallicRoughness.png" "https://www.web3d.org/x3d/content/examples/X3dForAdvancedModeling/GltfSampleModels/textures/lambert2_metallicRoughness.png"') [/X3D/Scene/Transform/Transform/Transform/Transform[1]/Transform/Group/Shape/Appearance/PhysicalMaterial/ImageTexture[4], warning]
- provided DEF, USE ='MetallicRoughness' while preserving containerField values
Gave it a white background for better visibility.
Added identifying WorldInfo.
When I ran the model through X3DJSAIL to reduce precision from 15 digits to 8 digits after decimal point, file size went from 867KB to 471KB. I think we can safely go down to 6 digits (micrometer resolution). Such reduction is valuable for large models, converters can have a switch for lossless/lossy conversion as an option.
Now online at
- X3D Example Archives: X3D4AM, X3D for Advanced Modeling, Gltf Sample Models, Cat By Muru Scene
- Converted glTF model showing use of Tangent node
- https://www.web3d.org/x3d/content/examples/X3dForAdvancedModeling/GltfSampleModels/CatByMuruSceneIndex.html
Castle and X_ITE look similar, X3DOM looks quite different (likely differing application of PBR textures).
<image.png>
<image.png>
<image.png>
all the best, Don-- X3D Graphics, Maritime Robotics, Distributed SimulationRelative Motion Consulting https://RelativeMotion.info
On Fri, Jan 2, 2026 at 5:05 PM Michalis Kamburelis <michalis.kambi at gmail.com> wrote:
I provided 3 sample models using Tangent in my answer to Don in
thread "[x3d-public] Specification review progress: PNG3, Tangent
node, HAnim weekly progress" on July :)
See https://github.com/castle-engine/demo-models/tree/master/bump_mapping/tangents
. It's probably easiest to "git clone" the whole repository
https://github.com/castle-engine/demo-models to have all additional
files (like textures).
Details:
- They look pretty and have been created by actual artists :) Original
versions obtained in glTF format from Sketchfab.
- They are on versions of Creative Commons licenses. So it's OK to
reuse them, if you credit the authors. 2 of them on CC-BY-4.0 (so even
commercial usage is allowed, you just need to credit the authors). 1
of them on CC-BY-NC-SA-4.0, so no commercial usage allowed (don't know
if it's a problem for X3D Examples Archive).
- They were created by converting glTF versions to X3D using Castle
Model Viewer.
- I removed some CGE extensions from them (they don't contain
animation in any way, including Skin), but admittedly I didn't test
them in other browsers. Reports about compatibility with other X3D
browsers are welcome. If there's anything non-spec-complaint in these
examples -> let me know and I'll fix it.
- Note that the models only make sense together with all the textures,
they have textures in "textures/" subdirectory.
- They are also larger files. I recall I tried to create some "more
minimal" test, but from what I recall they just looked ugly /
non-sensical :)
Regards and have a great 2026 everyone ! :),
Michalis
pt., 2 sty 2026 o 07:33 Don Brutzman <don.brutzman at gmail.com> napisał(a):
>
> happy 2026! i hope everyone is well.
>
> FYI i have deployed updates and support for draft x3d 4.1 development and will announce soon...
>
> X3D Specifications: Schema and DOCTYPE Validation
> https://www.web3d.org/specifications
>
> I was looking back through prior mail and had trouble finding some examples...
>
> If any of you have a simple X3D example for FontLibrary, or for Tangent node, I'd be happy to add it to our X3D Examples Archive. This will help me test/confirm all the tools are working OK.
>
> Appreciate your considering this request, and all your many efforts.
>
> all the best, Don
> --
> X3D Graphics, Maritime Robotics, Distributed Simulation
> Relative Motion Consulting https://RelativeMotion.info
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://web3d.org/pipermail/x3d-public_web3d.org/attachments/20260111/9fbb9621/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: SheenCloth.zip
Type: application/zip
Size: 6509122 bytes
Desc: not available
URL: <http://web3d.org/pipermail/x3d-public_web3d.org/attachments/20260111/9fbb9621/attachment-0001.zip>
More information about the x3d-public
mailing list