poppygl
Render GLTF files to PNG images in completely native JavaScript without WebGL/OpenGL.
Quick start
Give poppygl a .gltf or .glb URL and it will fetch every referenced buffer/texture, rasterize the scene, and hand back a PNG buffer.
import { renderGLTFToPNGBufferFromURL } from "poppygl"
import { writeFile } from "node:fs/promises"
const png = await renderGLTFToPNGBufferFromURL(
"https://models.babylonjs.com/DamagedHelmet.glb",
{
width: 800,
height: 600,
ambient: 0.15,
},
)
await writeFile("DamagedHelmet.png", png)
The fetch helper relies on the global
fetchAPI (present in Node 18+ and modern browsers). Pass a customfetchImplif you need different transport or caching behaviour.
Render from an in-memory GLB buffer
Already have the .glb bytes (for example, uploaded by a user or read from disk)? Send the buffer straight to renderGLTFToPNGBufferFromGLBBuffer and receive the PNG output.
import { readFile } from "node:fs/promises"
import { renderGLTFToPNGBufferFromGLBBuffer } from "poppygl"
const glb = await readFile("DamagedHelmet.glb")
const png = await renderGLTFToPNGBufferFromGLBBuffer(glb, {
width: 1024,
height: 768,
})
// write or return the PNG buffer
This helper expects every referenced buffer/image to be embedded in the GLB (the usual single-file package). If the asset links out to external resources, load it with
renderGLTFToPNGBufferFromURLinstead so poppygl can fetch the extras.
Render options
renderGLTFToPNGBufferFromURL accepts the same render options as the lower-level APIs:
width/height(default512): output resolution in pixels.fov: vertical field of view in degrees (defaults to35).camPosandlookAt: override the auto-framed camera position and target.lightDir: normalized directional light vector (defaults to a top-right key light).ambient: ambient lighting contribution (0–1, defaults to0.2).gamma: gamma correction applied to the output (defaults to2.2).cull: back-face culling mode ("back","front", or"none").fetchImpl: optional override for resource loading (must match thefetchsignature).
You can inspect the defaults via getDefaultRenderOptions() or reuse the internal merge logic with resolveRenderOptions().
If you already have the GLTF JSON
When the GLTF JSON object is already in memory (for example, bundled with your app), skip the network loader and supply resources directly:
import {
bufferFromDataURI,
createSceneFromGLTF,
decodeImageFromBuffer,
encodePNGToBuffer,
pureImageFactory,
renderSceneFromGLTF,
} from "poppygl"
import gltfJson from "./CesiumMan.gltf.json" assert { type: "json" }
import { readFile } from "node:fs/promises"
const base = new URL("./CesiumMan/", import.meta.url)
const buffers = await Promise.all(
(gltfJson.buffers ?? []).map(async (entry) => {
if (!entry.uri) throw new Error("Buffers without URIs need custom handling.")
return entry.uri.startsWith("data:")
? bufferFromDataURI(entry.uri)
: await readFile(new URL(entry.uri, base))
}),
)
const images = await Promise.all(
(gltfJson.images ?? []).map(async (img) => {
if (!img.uri) throw new Error("Only URI-backed images are shown in this example.")
const data = img.uri.startsWith("data:")
? bufferFromDataURI(img.uri)
: await readFile(new URL(img.uri, base))
return decodeImageFromBuffer(data, img.mimeType)
}),
)
const scene = createSceneFromGLTF(gltfJson, { buffers, images })
const { bitmap } = renderSceneFromGLTF(scene, { width: 512, height: 512 }, pureImageFactory)
const png = await encodePNGToBuffer(bitmap)
The only contract is that buffers is an array of Uint8Array instances and images is an array of BitmapLike textures (PNG and JPEG are supported out of the box via decodeImageFromBuffer).
Additional utilities
loadGLTFWithResourcesFromURLreturns{ gltf, resources }if you prefer to inspect or cache the parsed data before rendering.createSceneFromGLTFbuilds draw calls ready for the software rasterizer.computeSmoothNormalsandcomputeWorldAABBexpose useful preprocessing helpers.pureImageFactoryallocates thepureimagebitmap implementation used by the renderer.encodePNGToBufferpacks anyBitmapLikeinto a PNG buffer that can be written to disk or served over the network.
Happy rendering!