FBX2glTF

The --embed flag allows you to create a single distributable glTF file without using the binary .glb format. It achieves this by encoding the binary buffer(s) as base64-encoded data:// URIs within the JSON.

Note: This method is slow and results in much larger files compared to the .glb format, but it is useful for loaders that do not support .glb files.

FBX2glTF uses a "baking" method for animations. It steps through the animation's time interval keyframe by keyframe, calculates the local transform of each node, and records any changes in rotation, translation, or scale.

Key features and behaviors:

• Completeness: Every animation in the FBX file is converted to an animation in the glTF file.
• Blend Shapes: Supported on a per-mesh basis; clips can vary the weights associated with each shape.
• Trade-offs: Baking is simple and precise but can lead to very large file sizes, especially with complex animation rigs due to the data explosion of recorded keyframes.

FBX2glTF can optionally apply Draco compression to geometric data (vertex indices, positions, normals, per-vertex color, etc.). This is highly effective for reducing file size, particularly for static models.

Requirements:

• Enabling this adds a dependency on the KHR_draco_geometry_compression glTF extension.
• The target glTF viewer must be able to decompress Draco data.

The conversion process maps FBX material models (primarily Lambert and Phong) to glTF 2.0's Physically-Based Rendering (PBR) model. Because these models are inherently incompatible, some settings transfer automatically while others are handled differently.

Automatically supported settings:

• Emissive constants and textures
• Occlusion maps
• Normal maps

Traditional settings (Lambert/Phong) mapping notes: Settings like Ambient, Diffuse, Specular, and Shininess (which can be constants or textures) do not have a direct 1:1 mapping in PBR, as PBR relies on fundamental attributes like roughness and metalness.

glTF uses a coordinate system where (0, 0) is the top left, whereas FBX often assumes (0, 0) is the bottom left.

• --flip-v (the default behavior) applies y -> (1.0 - y) to all V texture coordinates to ensure spec-compliant glTF.
• --no-flip-v disables this behavior, which is useful if your textures are already pre-flipped or if you are working in a glTF-centric coordinate system.
• --flip-u applies x -> (1.0 - x) to all U texture coordinates.

If an FBX contains superfluous vertex attributes, you can use the -k or --keep-attribute flag to reduce the output file size. When this flag is used, you must specify it repeatedly for every attribute you wish to retain.

Supported arguments:

• position
• normal
• tangent
• binormial
• color
• uv0
• uv1

FBX2glTF is a command-line tool used to convert Autodesk FBX 3D models into the glTF 2.0 format. You can run it by providing the path to an FBX file as a positional argument, or by using explicit flags for input and output files.

# Simple conversion
> FBX2glTF ~/models/butterfly.fbx

# Complex pipeline with specific input, output, and compression
> FBX2glTF --binary --draco --verbose \
          --input ~/models/source/butterfly.fbx \
          --output ~/models/target/butterfly.glb

When using the convert function, you can control whether the output is a GLB (binary) or glTF (JSON) file using the opts array:

1. Explicit GLB: Provide a destination filename ending in .glb. If you also include --binary or -b in opts, the tool ensures consistency.
2. Explicit glTF: Provide a destination filename ending in .gltf.
3. Automatic GLB: If you include --binary or -b in the opts array, the function will automatically ensure the destination extension is .glb.
4. Automatic glTF: If no extension is provided for destFile, it defaults to .gltf.

FBX2glTF uses specific resolvers to map FBX materials to the glTF Metallic-Roughness model. The FbxRoughMetMaterialInfo structure holds the extracted properties, including baseColor, metallic, roughness, and emissive values, along with their corresponding texture maps (e.g., texNormal, texBaseColor, texMetallic).

Two primary resolvers handle this mapping:

1. FbxStingrayPBSMaterialResolver: Resolves Stingray PBS materials from FBX files.
2. Fbx3dsMaxPhysicalMaterialResolver: Resolves 3ds Max Physical materials from FBX files.

Both resolvers output an FbxRoughMetMaterialInfo object containing the processed material data and texture pointers.

You can run FBX2glTF using Docker Compose. The provided docker-compose.yaml defines a service named fbx2gltf that builds from the current directory context.

version: '3.7'
services:
  fbx2gltf:
    build:
      context: .