version int64 | dataset string | generated timestamp[s] | count int64 | rows list |
|---|---|---|---|---|
1 | StarAtNyte1/aether-assets | 2026-10-29T00:00:00 | 1,460 | [{"id":"aircon-exterior","name":"Outdoor air-conditioning unit","aliases":[],"tags":["aircon","exter(...TRUNCATED) |
Aether assets
1460 small, textured 3D models (GLB, real-world metres, base on y = 0, meshopt-compressed) for browser games and scenes, with a search index. They are the browser-sized models of the Aether asset library, published so that any world can find and load one lazily, without a local server.
Licences
This dataset has no single licence. Each asset keeps the licence of its source, shown in the licence column of index.json. Only two licences are accepted: CC0-1.0 (570 assets) and CC-BY-4.0 (890 assets). Nothing non-commercial, no-derivatives, share-alike or unlabelled is included.
- CC0 1.0 (legal code): no rights reserved. A credit is given anyway.
- CC BY 4.0 (legal code): use freely, credit the author, link the licence, and say that changes were made. Every CC BY asset below was modified (converted, reduced and recompressed by Aether); the
creditfield of each row is a ready attribution line.
Sources
Models derived from NVIDIA SimReady Central packages were converted by Aether from OpenUSD to GLB (world transforms baked, triangulated, welded, Z-up to Y-up, in metres), reduced in triangles and textures, and compressed. They are derivatives of the pinned source package (dataset, commit and file in source, with the source file's SHA-256 in models/<id>/provenance.json). They are not SimReady packages and have not been validated by NVIDIA.
Layout
models/<id>/<id>.glb main model (at most 1.2 MB)
models/<id>/<id>.lod1.glb fewer triangles, same pivot and scale (when present)
models/<id>/<id>.lite.glb smaller textures, for low-end devices (when present)
models/<id>/thumb.webp 192 px preview
models/<id>/provenance.json pinned source package (SimReady-derived models)
index.json one row per asset: id, name, aliases, tags, category, style, size (m), triangles, bytes,
licence (SPDX), author, credit, source (dataset, revision, path), conversion note,
collider, SHA-256 of every file
index.min.json the compact index the client loads (search only needs this)
credits.csv attribution table of every asset (author, licence, credit line, pinned source)
embeddings.bin + .json int8 CLIP ViT-B/32 image embeddings of the thumbnails (similar-looking models)
1460 models in 15 categories, 367.5 MB in all. Categories: props 655, furniture 161, kitchen 127, tools 105, decor 66, street 60, building 58, vehicles 39, nature 37, lighting 35, rocks 28, plants 25, food 24, animals 24, bathroom 16.
Use
from huggingface_hub import hf_hub_download
import json
index = json.load(open(hf_hub_download("StarAtNyte1/aether-assets", "index.json", repo_type="dataset")))
row = next(r for r in index["rows"] if "chair" in r["tags"])
glb = hf_hub_download("StarAtNyte1/aether-assets", row["files"]["main"]["path"], repo_type="dataset")
Check the file against files.main.sha256 before use. In a browser, fetch https://huggingface.co/datasets/StarAtNyte1/aether-assets/resolve/<commit>/index.min.json and search it locally; pin the commit.
Attribution (CC BY 4.0 assets)
890 CC BY 4.0 assets: the full table (asset, author, licence, ready-made credit line, source dataset and commit) is credits.csv. index.json carries the same fields per asset.
CC0 assets are listed in index.json with their author and source; the Kenney kits and Poly Haven models are public-domain dedications.
Safety
The files are data. The index hashes (SHA-256) let a client check every download. Loaders should still validate the GLB and cap its size.
- Downloads last month
- 49