Open-source platform for 3D AI agents. Turn text or a photo into a rigged, animated GLB avatar, give it an LLM brain, memory and a wallet, and embed it anywhere with one web component. Remote MCP server, x402 payments, Three.js, browser-native.
Bring your avatar: landing copy and post set (draft for owner approval)
Status: draft. Nothing here is published or posted. The owner approves the wording, then posts it. Wording is neutral on purpose: it names no company and makes no comparison, so it stays true whichever avatar service a reader came from.
Audience
People whose avatar lived in a cross-game avatar service that has since closed its hosted creator and public endpoints. Their exported .glb and .vrm files still load anywhere, but the apps and embeds that depended on the hosted service are left without a home for them. These readers already own a file. They need a place where that file keeps working and gains a purpose.
What is true today (the only facts the copy may use)
Every line below is shipped and covered by a test or a live page. Sources: import page, src/game/vrm-loader.js, src/game/vrm-expressions.js, src/animation-retarget.js, src/glb-canonicalize.js, tests/vrm-real.test.js, tests/import-rpm-source.test.js, Avatar engines, Embedding.
| Claim | Ground truth |
|---|---|
| Import by URL or by file | /import/rpm has a URL tab and a file tab. The server fetches URLs, so cross-origin rules do not block it. |
.glb and .vrm accepted |
The file tab accepts both. VRM 0.x and 1.0 load through the real @pixiv/three-vrm plugin. |
| VRM features work | MToon materials, spring-bone physics, node constraints, look-at and the 18 expression presets (tests/vrm-real.test.js). |
.vrma animation clips play |
Loaded through VRMAnimationLoaderPlugin and retargeted, so one clip plays on a VRM and on a Mixamo-named humanoid. |
| Voice and face capture drive VRM avatars | Expression presets map onto the ARKit-52 driver (src/game/vrm-expressions.js). |
| Any humanoid rig animates | Bone names from Mixamo, Avaturn, Unreal, VRM/VRoid, Daz, MakeHuman, Blender and simple rigs map to one canonical set, then the idle and walk clip library applies. A rig that cannot be skeleton-driven falls back to the default rig, never a T-pose. |
| The avatar is yours | It is stored in your account as a .glb, ready to become an agent with a wallet, and embeds anywhere with <agent-3d>. |
| Import reliability | A bug that made every import fail at the final save step was fixed with a regression test (tests/import-rpm-source.test.js). |
Do not claim: any count of imported avatars, any migration tool for a specific service, partnership or endorsement language, or any statement about another company's business.
Landing copy
Headline: Your avatar file keeps working here.
Subhead: Bring the .glb or .vrm you already have. It loads, moves and talks on three.ws, and you can put it on any website with one tag.
Primary button: Import my avatar (links to /import/rpm)
Three points
- Bring it as it is. Paste the file's URL or upload it. We read VRM 0.x and 1.0 with full support for toon materials, hair and cloth physics and facial expressions.
- It moves right away. Whatever rig it has, it plays the idle and walk library. VRM animation clips (
.vrma) play too, and voice or camera capture can drive its face. - Give it a job. Turn it into an agent with its own page and Solana wallet, or embed it on your site:
<script type="module" src="https://three.ws/agent-3d/latest/agent-3d.js"></script>
<agent-3d src="https://example.com/my-avatar.glb"></agent-3d>
How to find your file: if your app loaded the avatar from a link, open the browser's developer tools, switch to the Network tab, filter by .glb, and copy the request URL. If you saved the file earlier, use the file tab.
Footer line: Free to import. Your file stays yours.
Implementation note: the existing page pages/import-rpm.html already is the destination and is marked noindex. Publishing this copy as a public page means adding a data/pages.json entry, removing noindex, and a deploy, which are owner-gated. The wording above is ready to drop into that page.
Post set (not posted)
Each post links to https://three.ws/import/rpm. Add the image from a real import screenshot taken by the owner. Tone: friendly opening, technical detail after.
Post 1, launch (X) If you have an avatar file sitting on your drive and nowhere to use it, bring it to three.ws. Paste a link or drop a .glb or .vrm and it moves, talks and embeds on any site. Free to import.
Post 2, VRM depth (X) VRM support on three.ws is the real thing: MToon toon shading, spring-bone hair and cloth, 18 expression presets and .vrma animation clips. Same clip, any humanoid rig. Voice and camera capture drive the face.
Post 3, one tag (X)
Put your avatar on your website with one tag:
<agent-3d src="your-avatar.glb"></agent-3d>
No account needed to embed. Import yours at three.ws/import/rpm.
Post 4, make it an agent (X) An avatar is better with a purpose. Import it, give it a personality, and it becomes an agent with a public page and its own Solana wallet. $THREE holders and builders are already doing this.
Post 5, how-to thread (X, 4 parts)
- Find your avatar's file: open dev tools, Network tab, filter by .glb, copy the URL.
- Open three.ws/import/rpm and paste it, or upload the file.
- Pick a name. Your avatar is saved to your account.
- Embed it, animate it, or turn it into an agent.
Long-form article outline (blog or community forum)
- Friendly opening: what it feels like to own an avatar file and want it to keep working.
- What import accepts and where the file ends up.
- VRM in depth: shading, physics, expressions,
.vrma, one clip across rigs (technical middle). - Embedding with
<agent-3d>and driving moods from page code. - Turning an avatar into an agent: page, wallet, skills.
- Ecosystem and partners: list only partner designations already published in the repo's partner docs, with their exact wording.
- Technical appendix: bone canonicalization, retargeting, how expressions map to ARKit-52.
Owner step
Approve this wording and post it. Publishing the public page is the single command-level step: add the data/pages.json entry, drop noindex from pages/import-rpm.html, and deploy. Nothing else is outstanding.