About VRChat World Search

A fast, static search over public VRChat worlds: full-text search (including Japanese and other CJK text), filters for capacity, download size, platform, dates, tags and author, and "similar worlds" suggestions. It runs entirely in your browser from static files; there is no search server. Made by Hiina.

Data

The world list is public VRChat world metadata (name, description, author name, tags, capacity, sizes, dates, visit and favorite counts) collected by the author's own scraper. Only worlds whose release status is public are included (197,992 worlds). Private worlds are not.
Snapshot: newest scrape 2026-10-20. Individual worlds may have been scraped earlier, and anything changed on VRChat since then (renamed, made private, deleted) is not reflected until the next refresh.
This build has no thumbnails; the detail panel links to each world's image and official page on vrchat.com.
The underlying database is public: a live mirror, updated daily, is on DoltHub, and the scraper and this site's source are on GitHub.
This site is not affiliated with, endorsed by, or sponsored by VRChat Inc. "VRChat" is a trademark of VRChat Inc. World names, descriptions and images belong to their authors.

Removal requests

If you are a world author (or VRChat) and want a world or your whole author profile removed, or a thumbnail or text corrected, email [email protected] with the world id (wrld_..., or the vrchat.com link) or the author id (usr_...). You do not need to give a reason. Removals take effect at the next rebuild; I aim to do that within about a week. A world you made private or deleted on VRChat stays here until the next refresh, so email if you want it gone sooner.

Privacy

No cookies, no analytics, no third-party requests: everything the page loads comes from this site. Your searches are evaluated locally and never sent anywhere. The only thing stored is your preferences (for example, that you turned on semantic search) in your browser's localStorage. The host (Cloudflare Pages) keeps ordinary server logs of requests, as any web host does.

Credits and licences

Similar worlds: neighbours are precomputed offline with the text embedding model embeddinggemma-2 (Apache-2.0). Only the resulting neighbour lists ship with the site.
Semantic search (optional, off by default; downloads about 33 MB after you confirm): static-similarity-mrl-multilingual-v1 from the sentence-transformers project (Apache-2.0), quantised and truncated, run in your browser with a small JavaScript reimplementation of its tokenizer (licence text, changes and citations).
Code and fonts: the page is plain TypeScript built with Vite (MIT); it bundles no third-party libraries and no web fonts (your system fonts are used).

Author: Hiina, [email protected].