cstr/open-emoji-assets

Open Emoji Assets

Six permissively-licensed emoji sets in one place, with a single index that makes them searchable in eight languages.

Built because there is no coherent, redistributable, permissively-licensed emoji asset collection: the sets exist but are scattered across projects with different naming schemes, different coverage and different licences, and the obvious-looking alternatives are traps. The popular "Animated Fluent Emojis" repository is Personal Use Only; the stock libraries (LottieFiles, IconScout, Flaticon) let you use assets but forbid redistributing them as a set; and the existing emoji datasets on the Hub are ML training blobs — Arrow and Parquet — often with no licence declared, which is legally worse than the upstream repository they were copied from.

Sets

set	files	licence	format
noto-animated	884	CC-BY-4.0	lottie	animated
twemoji	2,970	CC-BY-4.0	svg	static
noto	3,682	OFL-1.1	svg	static
blobmoji	2,356	Apache-2.0	svg	static
emojitwo	1,840	CC-BY-4.0	svg	static
fluent	1,495	MIT	svg	static

No ShareAlike. No NonCommercial. Nothing here constrains how your own application is licensed — including dual AGPL/commercial. Two of the six are outright permissive (MIT, Apache-2.0); the rest are CC BY, which needs one credit line in an About screen, not per-item notices.

noto-animated is the only redistributable animated emoji set in existence, as far as we could establish. Lottie is used rather than GIF because it is roughly 20x smaller (~36-66 KB vs 190 KB-1.1 MB per emoji) and vector, so it stays sharp at any size.

index.json

One file describing all 3,944 Unicode emoji:

{
  "slug": "1f44d",
  "char": "\ud83d\udc4d",
  "name": "thumbs up",
  "group": "People & Body",
  "subgroup": "hand-fingers-closed",
  "annotations": {
    "de": { "name": "Daumen hoch", "keywords": ["Daumen", "einverstanden", ...] },
    "ja": { "name": "\u30b5\u30e0\u30ba\u30a2\u30c3\u30d7", "keywords": ["OK", "\u3044\u3044\u306d", ...] }
  },
  "search": ["+1", "daumen hoch", "good", "like", "\u3044\u3044\u306d", ...],
  "sets": { "noto-animated": "lottie", "twemoji": "svg", "fluent": "svg" }
}

• search is a flattened, lowercased, deduplicated list of every term across all locales — 236,977 terms in total, about 60 per emoji. One substring match over it is enough to power a picker's search box.
• sets lists only what is actually present in this repository; it is generated by reading the files off disk, so it can never point at something that was not shipped. Coverage genuinely differs between projects.

Metadata comes from Unicode's emoji-test.txt plus CLDR annotations and annotationsDerived for en, de, fr, es, it, nl, pt, ja. The derived file matters: without it, flags, skin-tone variants and ZWJ sequences have no keywords at all.

Layout

index.json
noto-animated/lottie/<slug>.json
twemoji/<slug>.svg
noto/<slug>.svg
blobmoji/<slug>.svg
emojitwo/<slug>.svg
fluent/<slug>.svg
licenses/
ATTRIBUTION.md

<slug> is the Unicode codepoint sequence, lowercase, underscore-separated (1f44d, 1f1e9_1f1ea, 1f636_200d_1f32b_fe0f) — consistent across every set, which the upstream projects are not.

Reproducing

build_set.py and enrich.py rebuild this from the upstream sources. Nothing is vendored blindly: which emoji have animations is not published anywhere, so the builder probes the CDN, and Fluent's paths are resolved from its repository tree because deriving them from Unicode names only matches about 60%.

Attribution

See ATTRIBUTION.md for per-set notices and suggested wording.