DAE File Documentation

---

Technical details

Feature	Value
Full name	COLLADA Digital Asset Exchange
File extension	.dae
MIME type	model/vnd.collada+xml
Format type	XML-based 3D scene/model interchange (plain text)
Developer	Khronos Group (originally Sony Computer Entertainment)
Introduced	2004 (COLLADA 1.0)
Latest version	COLLADA 1.5.0 (2008); ISO/PAS 17506
Open standard	Yes
Base format	XML
Root element	<COLLADA> with a collada.org schema namespace
Stores	Meshes, materials, textures, scene graph, cameras, lights, rigs, animation
Textures	Referenced by path, not embedded
Modern successor	glTF / GLB (also from Khronos)
Open with	Blender, SketchUp, macOS Preview/Quick Look, online 3D viewers
Common conversions	OBJ, FBX (other 3D tools); glTF, GLB (web/AR)
Related extensions	.obj, .fbx, .gltf, .glb, .3ds, .stl, .kmz
Specification	khronos.org/collada

What is a DAE file?

A .dae file is a COLLADA 3D model. COLLADA stands for COLLAborative Design Activity, and .dae for Digital Asset Exchange, which is the format’s whole purpose: to carry a 3D scene from one program to another regardless of vendor. It was created by Sony Computer Entertainment for PlayStation tooling and released in 2004; since 2006 it has been maintained by the Khronos Group, the same standards body behind OpenGL, Vulkan and glTF. The last revision, COLLADA 1.5.0, arrived in 2008.

A DAE file is plain-text XML that describes a complete 3D scene: meshes (vertices and polygons), materials and shading, texture references, a scene-graph node hierarchy, cameras, lights, skeletal rigs and keyframe animation. Because it is text and schema-defined, almost any 3D application can import or export it, which is exactly why it became a common interchange format, notably for SketchUp models and for Google Earth (which packages DAE inside a KMZ). Its MIME type is model/vnd.collada+xml. Being XML, it can be opened in a text editor, but only a 3D viewer renders the model.

The COLLADA root and library structure

A DAE file opens with an XML declaration and a single <COLLADA> root element whose namespace URL identifies the schema version. Everything inside is organised into library_* blocks, each a collection of one kind of asset, plus a <scene> that instantiates them. This separation of libraries from the scene is the format’s central idea: geometry, materials and images are defined once, then referenced by the scene graph.

<?xml version="1.0" encoding="utf-8"?>
<COLLADA xmlns="http://www.collada.org/2005/11/COLLADASchema" version="1.4.1">
  <asset> ... units, up-axis, authoring tool, dates ... </asset>
  <library_images>        ... external texture file references ...
  <library_effects>       ... shader/effect definitions ...
  <library_materials>     ... materials that instance effects ...
  <library_geometries>    ... meshes: positions, normals, UVs, polygons ...
  <library_visual_scenes> ... node hierarchy: transforms, instances ...
  <scene>                  ... which visual_scene to render ...
</COLLADA>

The <asset> block is worth noting because it carries scene-wide settings that cause real-world import problems: the measurement unit (metres, centimetres, inches) and the up_axis (Y-up or Z-up). Different tools default differently, which is why a COLLADA model sometimes imports at the wrong scale or lying on its side, and why the <asset> metadata exists to make those choices explicit.

library_geometries: how a mesh is stored

Mesh data lives in <library_geometries>, and COLLADA stores it in a deliberately indirect way built from <source>, <vertices> and a primitive element such as <triangles> or <polylist>. A <source> holds a flat <float_array> of raw numbers, positions, normals or texture coordinates, together with an <accessor> that says how to read them (for example, in strides of three for XYZ). The primitive element then references those sources through <input> channels with a semantic (POSITION, NORMAL, TEXCOORD) and an offset, and a <p> element lists the interleaved indices that assemble the actual faces.

This indexed, source-plus-accessor design is more verbose than a format like OBJ, but it lets one array of positions be shared by several primitive sets with different normals or materials, and it makes the format self-describing about how each stream is laid out. It also explains why DAE files are large and slow to parse compared with binary 3D formats: every number is ASCII text inside XML.

Materials, effects and why models arrive grey

Shading is split across two libraries. <library_effects> defines the actual shader parameters (a Phong or Lambert profile, colours, texture samplers), and <library_materials> defines named materials that each instance one effect. The scene’s geometry then binds a material to a set of faces by name. Crucially, <library_images> stores only references to texture image files, their file paths, not the image data itself. This is the single most common practical problem with DAE files: because the textures are external, a model shared as a lone .dae arrives grey or untextured when the accompanying image files were not included or their paths broke. The fix is to ship the texture files alongside the DAE, or to re-link them in the 3D application after import.

The scene graph, rigs and animation

The <library_visual_scenes> block holds the node hierarchy: a tree of <node> elements, each with transforms (<translate>, <rotate>, <scale> or a full <matrix>) and instances that place geometry, cameras or lights into the world. Optional libraries add motion: <library_animations> stores keyframe channels that drive node transforms over time, and <library_controllers> stores skinning data (a skeleton, bind poses and per-vertex bone weights) for character rigs. Not every DAE uses these; many are static props with only geometry and materials. This matters for conversion, because simpler target formats silently drop the animation and rig, as the next section describes.

Converting DAE to OBJ, FBX and glTF

People rarely keep a model as COLLADA; they convert it for a specific pipeline, and the choice of target determines what survives. Import the .dae into Blender (File › Import › Collada), then export to the format you need.

To OBJ gives a simple, near-universal mesh, but OBJ is static geometry only, so any animation and rigging in the DAE is lost. To FBX is the route into game engines such as Unity and Unreal and keeps materials plus animation and rigging that OBJ drops. To STL keeps only the raw triangle mesh for 3D printing, discarding colour and materials. The most important modern conversion is to glTF or GLB: glTF is Khronos’s newer successor to COLLADA, leaner and binary-friendly, and it is the standard for web 3D (three.js) and AR (Apple’s AR Quick Look). GLB is the single-file binary form of glTF, packing geometry and textures together, which makes it the easiest format to share. Converting DAE to glTF/GLB is a clean, well-defined step because both formats come from the same body and cover the same scene concepts. One conversion to avoid is DAE to STEP or STP: COLLADA is a polygon mesh while STEP is precise CAD solid geometry, and there is no faithful mesh-to-solid conversion, so go back to the original CAD source if you need that.

Opening a DAE without a full 3D suite

You do not need expensive software to view a COLLADA model. Blender (free, all platforms) imports it for full editing and is the best general option. SketchUp handles COLLADA directly, since it is SketchUp’s standard interchange format. On macOS you can simply press Space on the file for a Quick Look preview, or open it in Preview, both of which render COLLADA natively. In a browser, drag the .dae into an online viewer such as 3dviewer.net to rotate and inspect it with no install. One correction worth stating plainly: older references claim Adobe Photoshop and AutoCAD open DAE files. Photoshop’s 3D engine, which once imported COLLADA, was deprecated and its 3D features removed in 2024, and AutoCAD is CAD software rather than a mesh viewer, so neither is an appropriate current DAE tool.

References

• Khronos Group — COLLADA overview
• Blender Manual — Collada (.dae) import and export
• Library of Congress — COLLADA / DAE format

Feedback

Was this page helpful?

Yes No

Glad to hear it! Please tell us how we can improve.

Sorry to hear that. Please tell us how we can improve.