FBX to DAE converter — into an older pipeline

VTubeMe's FBX to DAE converter writes your .fbx as a Collada 1.4.1 file with the skeleton, skin weights and shape keys, zipped with its base-colour textures. Free, five a day, up to 50 MB. Animation does not travel.

Free · 5 conversions a day per accountNeed a VRM avatar instead? →

How do I convert an FBX to DAE (Collada)?

Drop the .fbx on this page and download a zip holding the .dae and its textures.

  1. 1Drop your binary .fbx into the box above, up to 50 MB; sign in when prompted.
  2. 2The file is opened in Blender on our server and written with its Collada exporter: triangulated meshes, skeleton, skin weights, shape keys, and each material's base-colour image beside the .dae.
  3. 3Download the zip and keep its files in one folder.
Accepts:
one binary .fbx (7.1 or newer) up to 50 MB.
Returns:
a zip: one Collada 1.4.1 .dae (Z-up, metres) and its base-colour textures.
Costs:
nothing, and no credits — an account, though: five conversions per account per rolling 24 hours, shared across all the free format pairs. Conversions that fail are not counted.
Measured:
73 bones, 10 skinned meshes, 118 shape keys and 10 materials in, the same back in Blender; 8 of the FBX's 15 images written.
  • Only base-colour textures are written.
  • Shape keys make the .dae large — about 29 MB for the test character.
  • Verified by re-import in Blender only; animation does not travel.

What actually happens to the file

FBX is the current interchange format and Collada the previous one. You go backwards for a reason: Second Life and OpenSim take rigged mesh as .dae, SketchUp imports it, and some simulation and older engine toolchains were never updated.

On a rigged character the Collada file reads back in Blender with the same 73 bones, 10 skinned meshes, 118 shape keys, 10,977 vertices and 16,867 triangles, at the same height. The FBX's centimetres become Collada's declared metres.

Fewer textures arrive than left: the FBX embedded 15 images and the zip holds 8, the base-colour ones.

What this conversion cannot do

Measured on real files, not assumed. Everything below is a property of the formats or of the pipeline, and knowing it beforehand is cheaper than discovering it in your 3D tool.

Animation does not travel

The skeleton is written, the motion curves are not. Reading an FBX's animation costs about 1.5 GB of memory on a long clip, which is more than the converter container has, so the importer is asked to skip it and the exporter writes none. A file that arrives with a walk cycle leaves with the same rig in its bind pose. If what you have is the motion rather than the model, the FBX to VRMA converter is the tool that keeps it.

Base-colour textures only

The download is a zip: the .dae and the images its materials name, by relative path. On the sample avatar that was 8 of the 15 images — exactly the eight bound to a base colour. Normal maps, roughness and alpha maps are not written, because the Collada writer exports a material's diffuse texture and nothing else. The eight are written in their own format, a JPEG staying a JPEG.

Shape keys make the file very large

Collada is XML, and every shape key is written out as a complete second copy of its mesh in text. The sample avatar — 16,867 triangles, 118 shape keys — produced a .dae of 29.6 MB from a 3 MB model. A model with no shape keys does not have this problem. If the destination ignores morph targets, as most old Collada importers do, an FBX or GLB will serve you better.

Read back in Blender, and nowhere else

The result was re-imported with Blender's Collada reader: the counts above came back and the skinned mesh measured the same 1.7605 × 0.3017 × 1.8143 m. It was not opened in SketchUp, Second Life, OpenSim, Godot, Assimp-based tools or any older engine, and Collada is the format where importers disagree most. The file is Collada 1.4.1, Z-up, one unit to the metre, triangulated.

Uploads stop at 50 MB

This pair takes files up to 50 MB — half of what the GLB, FBX and OBJ pairs take among themselves, because dense scans and print files cost the converter far more memory per megabyte than a game character does. For scale: a binary STL of 50 MB is about a million triangles, and we converted one of 999,698. If your file is larger, decimate it first; MeshLab and Blender both do that without changing the format.

Five conversions a day, per account

This pair is free and takes no credits, but one account gets five conversions per rolling 24 hours across all the free format pairs together — the converter is one machine and this is what keeps it answering. Each slot frees itself 24 hours after it was used; nothing expires at midnight. Conversions that fail are not counted.

Frequently asked questions

Will this work for a Second Life or OpenSim mesh upload?

It is the right container — both take Collada 1.4.1 with a skeleton — but we did not test an upload, and those platforms have their own rules about bone names, joint counts and triangle limits that a conversion does not check. Treat the file as a starting point.

Why is the download a zip?

Because a textured .dae is several files: the XML document and the images it names. Unpack the zip and keep them in one folder — the .dae finds its textures by relative path, so moving it alone gives you a grey model.

When is Collada still the right target?

When the program on the other end takes little else: SketchUp, Second Life and OpenSim mesh upload, older game engines and toolchains from the late 2000s, some robotics and simulation packages. For anything current — Unity, Unreal, Godot 4, three.js — FBX or GLB is the better-supported road.

Can I preview the .dae?

Yes — the DAE viewer opens it in your browser; drop the .dae together with its images. Open the DAE viewer →

The other format pairs

All 30 are free, all run the same pipeline, and all say what they drop.

Only want to look at the file? GLB viewer, FBX viewer and OBJ viewer all run on your own device — nothing is uploaded.

Is your model going into VTubing?

If the model is a rigged character and it is headed for VTubing, none of these containers is the destination — VSeeFace, Warudo and VNyan load VRM. The VRM converter takes the same rigged FBX or GLB, maps its skeleton onto the VRM humanoid and converts the materials.

Open the VRM converter