Error: GLB must have 0–1 buffers. when writing glb

[daniel-gato]

I'm trying to convert a gltf to glb.

The gltf has this specificities:

• uses draco
• uses texture transform
• the textures and binaries are remote urls
• the gltf references many .bin and .jpg

I keep getting this message:

• Error: GLB must have 0–1 buffers.

I don't have control over that gltf and it displays correctly on gltf viewers.

I'm not sure why this happens. I would expect the write() to include all .bins into the file.

import fetch from 'node-fetch';
import { NodeIO } from '@gltf-transform/core';
import { KHRONOS_EXTENSIONS } from '@gltf-transform/extensions';

const io = new NodeIO(fetch)
  .setAllowHTTP(true)
  .registerExtensions(KHRONOS_EXTENSIONS);

const doc = await io.read('/tmp/model.gltf');
await io.write('/tmp/model.glb', doc);

Am I missing something here?

[donmccurdy]

Hi @daniel-gato! By convention a .glb is usually self-contained (i.e. does not reference external .bin files) and so any buffers need to be consolidated if there are >1. The I/O classes don't currently do that, in order to avoid mutating the original Document during export unintentionally.

If you add the following code before writing the Document, it will consolidate any extra buffers:

const buffer = doc.getRoot().listBuffers()[0];
doc.getRoot().listAccessors()
  .forEach((a) => a.setBuffer(buffer));
doc.getRoot().listBuffers()
  .forEach((b, index) => (index > 0 ? b.dispose() : null));

EDIT: In more recent versions of the library, this shortcut can be used instead:

import { unpartition } from '@gltf-transform/functions';

await document.transform(unpartition());

[daniel-gato]

I’m not sure to understand the code. Will that remove the extra buffers or merge them into one? I’m other terms, will the resulting gob be visually the same than the initial gltf? My understanding was that a gltf contains the binary geometry data into one or several external binary files. While the glob contains all those files and the gltf structure merged into one single binary file.

[donmccurdy]

Each Buffer is a flexible container for binary data, like mesh geometry or animation. Any number of vertex attributes or animation tracks can be stored in a single Buffer. To summarize what the code is doing:

// select the first buffer
const buffer = doc.getRoot().listBuffers()[0];

// reassign all binary data to the first buffer
doc.getRoot().listAccessors()
  .forEach((a) => a.setBuffer(buffer));

// delete all buffers except the first
doc.getRoot().listBuffers()
  .forEach((b, index) => (index > 0 ? b.dispose() : null));

There will be no visual differences in the GLB, but all of the external .bin (buffer) files will be stored in the file's single internal buffer. Only one buffer can be embedded in a .glb, so we have to merge their data.