Skip to content

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.

docs(changelog): log the Animation Studio edit tools and MoMask backend, update best3d progress

nirholas0f5bbfb46166
diff --git a/CHANGELOG.md b/CHANGELOG.md
index 76f84b285..dcf4632ec 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -13,8 +13,10 @@ Public history for [three.ws](https://three.ws), newest first. New pages come fr
- **Docs · Engine bridges** (`/docs/engine-bridges`): One Forge API contract, three real editor integrations: a Unity UPM package, a Godot addon, and an Unreal plugin, each generating a mesh from a prompt and importing it with the engine's own native glTF pipeline.
- **Docs · Engine QA** (`/docs/engine-qa`): A free real-time-engine fitness grade for any 3D model: draw calls, material count, texel density, a per-level LOD triangle budget, and whether the mesh is cheap and closed enough to use as its own physics collider.
- **<agent-3d> gets model-viewer's AR contract: ar-modes, variants, hotspots, and /ar drops the third-party CDN**: <agent-3d> now matches the AR/viewer contract the market expects from model-viewer (Apache-2.0): ar-modes picks which launch paths to try (WebXR, Scene Viewer, Quick Look), ar-placement anchors floor or wall, ios-src supplies a pre-baked USDZ or the element generates a best-effort one on device, variant-name switches a GLB's real KHR_materials_variants materials live, and any element with slot="hotspot" tracks a 3D position and flies the camera to a saved orbit on click. /avatars/:id/ar and /agents/:id/ar (pages/ar.html) now run entirely on <agent-3d ar>; the page no longer loads a third-party viewer from a CDN. (`/avatars`) `[feature]`
+- **Animation Studio: trim, loop-blend, and save any generated clip to your library**: The Animation Studio's prompt-to-clip flow (type a prompt, generate motion on the self-hosted GPU worker, preview it retargeted onto the loaded avatar) now has an editing and persistence half. The transport grows trim start/end inputs, a loop-blend duration field, and a loop toggle, all recomputed through a non-destructive edit path so the live preview, the GLB export, and the saved clip always agree on the current edit. A Save to library button collects a name and visibility and posts the canonical, pre-retarget clip to the shared clip library, so it plays back correctly on any future rig, not just the avatar it was previewed on. A separate bug that crashed every generated clip's playback the moment it reached the browser (AnimationClip.parse() expected userData as a JSON string, the server was sending an object) is also fixed. (`/pose`) `[feature, fix]`
- **Engine QA: a free real-time-engine fitness grade, plus real LOD chains**: A new free grade answers the question a game engine's import step asks that neither glTF validation nor simulation readiness does: draw calls, material count, texel density against the mesh's real-world size, a per-level triangle budget for a LOD chain, and whether the mesh is cheap and closed enough to use as its own physics collider. GET/POST /api/3d/engine-qa grades any public GLB (format=stl exports a watertight mesh for printing); the same browser-only checks run with zero network cost in Rig Doctor, and the full grade with a one-click upgrade in the GLB viewer, through a shared <engine-qa> panel component. The underlying LOD builder (packages/glb-tools's buildLodChain, real meshoptimizer decimation tagged with MSFT_lod) is now also available on generation: send { "lods": true } to POST /api/forge and the finished creation gets a combined LOD-chain variant alongside its delivered mesh, the same opt-in, fire-and-forget pattern as the existing web-delivery variant. (`/docs/engine-qa`) `[feature]`
- **Real VRM avatars: spring bones, expressions, and .vrma clips via @pixiv/three-vrm**: The hand-rolled VRM loader is gone. vrm-boot.js now registers the real VRMLoaderPlugin (@pixiv/three-vrm) on the three.ws GLTFLoader, so VRM 0.x and 1.0 avatars get genuine MToon materials, spring-bone physics, node constraints, look-at, and all 18 expression presets, driven live by vrm.update(delta) in AnimationManager and mapped onto the ARKit-52 face-capture driver. VRMAnimationLoaderPlugin loads .vrma clips into the same retargeting pipeline that drives non-VRM humanoids, so a single clip plays on either. The Ready Player Me import page (/import-rpm) now accepts .vrm files alongside .glb, and a long-standing bug that 400'd every import at the final save step (it sent an invalid source value) is fixed. docs/avatar-engines.md upgrades three-vrm from interop to live. (`/docs/avatar-engines`) `[feature, fix]`
+- **Text-to-motion gets a second GPU backend: MoMask, selectable per request**: The self-hosted text-to-motion worker behind the Animation Studio and POST /api/forge-motion now runs two independent generation backends on the same GPU lane: MDM (Motion Diffusion Model, MIT), the existing default, and MoMask (EricGuo5513/momask-codes, MIT), a residual-VQ tokenizer plus masked and residual transformer that publishes a lower HumanML3D FID (0.045 vs. T2M-GPT's 0.141). Pass model: "momask" to /infer or /api/forge-motion to route a prompt through it instead; omit it and MDM keeps answering as before. Both backends converge on the same (T,22,3) HumanML3D position format and the same shared decode math, so a MoMask clip previews, retargets, and exports exactly like an MDM one. The two projects ship conflicting top-level `utils` packages; the worker resolves that at load time so either backend can be warmed in the same container without the other's modules leaking in. MoMask's checkpoints are staged to production storage and the worker is wired to load them; the first on-GPU deploy and a real in-house benchmark against MDM are tracked as the next step. (`/pose`) `[feature, infra]`
- **Three new engine bridges: Unity, Godot, and Unreal, one prompt to a mesh in your editor**: A shared contract (specs/ENGINE_BRIDGE.md) and three real, independent implementations: a Unity UPM package that imports with glTFast, a Godot addon that imports with Godot's own native GLTFDocument, and an Unreal plugin that imports through Unreal's Interchange glTF pipeline. Each one calls only the public Forge API (POST /api/forge to submit, GET /api/forge?job=<id> to poll, a plain HTTPS GET to download), the same free-first routing and anonymous x-forge-client auth every other Forge caller gets, with the same four import settings (generate_colliders, import_materials, scale, destination) mapped onto each engine's own equivalent. A dropdown and a Generate button in each editor: type a prompt, get a mesh dropped into your scene. (`/docs/engine-bridges`) `[feature, sdk]`
## 2026-10-11
diff --git a/data/changelog.json b/data/changelog.json
index 461978ce0..bab24c90d 100644
--- a/data/changelog.json
+++ b/data/changelog.json
@@ -1,6 +1,26 @@
{
"$comment": "Curated changelog entries for three.ws, the editorial layer of the public changelog. New PAGE launches are tracked automatically via the `added` field in data/pages.json; this file is for everything else holders care about: improvements, fixes, SDK releases, trading/payments capabilities, security work. scripts/build-page-index.mjs merges both sources into CHANGELOG.md, public/changelog.json, and public/changelog.xml. Newest first. Allowed tags: feature, improvement, fix, sdk, infra, docs, security, desktop (desktop entries become the desktop app's release notes). `link` is optional and must be a live three.ws path.",
"entries": [
+ {
+ "date": "2026-10-12",
+ "title": "Text-to-motion gets a second GPU backend: MoMask, selectable per request",
+ "link": "/pose",
+ "summary": "The self-hosted text-to-motion worker behind the Animation Studio and POST /api/forge-motion now runs two independent generation backends on the same GPU lane: MDM (Motion Diffusion Model, MIT), the existing default, and MoMask (EricGuo5513/momask-codes, MIT), a residual-VQ tokenizer plus masked and residual transformer that publishes a lower HumanML3D FID (0.045 vs. T2M-GPT's 0.141). Pass model: \"momask\" to /infer or /api/forge-motion to route a prompt through it instead; omit it and MDM keeps answering as before. Both backends converge on the same (T,22,3) HumanML3D position format and the same shared decode math, so a MoMask clip previews, retargets, and exports exactly like an MDM one. The two projects ship conflicting top-level `utils` packages; the worker resolves that at load time so either backend can be warmed in the same container without the other's modules leaking in. MoMask's checkpoints are staged to production storage and the worker is wired to load them; the first on-GPU deploy and a real in-house benchmark against MDM are tracked as the next step.",
+ "tags": [
+ "feature",
+ "infra"
+ ]
+ },
+ {
+ "date": "2026-10-12",
+ "title": "Animation Studio: trim, loop-blend, and save any generated clip to your library",
+ "link": "/pose",
+ "summary": "The Animation Studio's prompt-to-clip flow (type a prompt, generate motion on the self-hosted GPU worker, preview it retargeted onto the loaded avatar) now has an editing and persistence half. The transport grows trim start/end inputs, a loop-blend duration field, and a loop toggle, all recomputed through a non-destructive edit path so the live preview, the GLB export, and the saved clip always agree on the current edit. A Save to library button collects a name and visibility and posts the canonical, pre-retarget clip to the shared clip library, so it plays back correctly on any future rig, not just the avatar it was previewed on. A separate bug that crashed every generated clip's playback the moment it reached the browser (AnimationClip.parse() expected userData as a JSON string, the server was sending an object) is also fixed.",
+ "tags": [
+ "feature",
+ "fix"
+ ]
+ },
{
"date": "2026-10-12",
"title": "Real VRM avatars: spring bones, expressions, and .vrma clips via @pixiv/three-vrm",
diff --git a/prompts/finish/_context/best3d-PROGRESS.md b/prompts/finish/_context/best3d-PROGRESS.md
index a8143aacc..361c44fb8 100644
--- a/prompts/finish/_context/best3d-PROGRESS.md
+++ b/prompts/finish/_context/best3d-PROGRESS.md
@@ -73,3 +73,12 @@ Append a dated entry per closed order, with commit SHAs.
- DoD line 4 (docs, STRUCTURE.md, changelog): all three already existed from the June 2026 ship. `STRUCTURE.md` row 541 (`@three-ws/voice`) names `api/a2f.js` and "audio2face lipsync visemes." `data/changelog.json` carries two user-facing entries (`"Avatars can move their lips again: the Audio2Face lane is back"` and `"Audio2Face: your avatar's face now moves with its voice"`, linking `/audio2face`). Docs run deep: `packages/voice/README.md` documents the full wire contract and client API, `packages/audio-mcp/README.md` exposes an `audio_to_face` MCP tool over the same lane, and a dozen `docs/*.md` files (nvidia-models.md, the NVIDIA forum write-ups, mcp-studio.md, the-first-19-weeks.md and others) describe it. Nothing to add; `npm run build:pages` not re-run since no changelog/pages content changed.
- No code touched this pass; this entry plus the order-file deletion are the only diff. No owner action remains: the lane needs no GPU deploy (it already runs against NVIDIA's own hosted NVCF endpoint, already configured in production), so no row was added to `production-100-OWNER-ACTIONS.md`.
- Commit: (this entry + `git rm prompts/finish/076-best3d-10-audio2face-worker.md`, see the commit this entry ships with).
+- 2026-10-12: order 077 (prompt-to-clip on any rig). Tasks 2 and 3 (Animation Studio prompt box, retarget preview, keyframe timeline with trim/blend/loop, save-to-library for any canonical rig) were already built and committed in an earlier session this same day, before this pass started: `1705530bc` (fix: stop `AnimationClip.parse()` crashing on every generated clip, `userData` needed to be a JSON string, not an object) and `58bd13cb9` (feat: trim start/end, loop-blend duration, loop toggle, a non-destructive `_computeEditedClip()`, and a Save to library modal posting the canonical pre-retarget clip to `/api/animations/clips`). This pass's own work was task 1 (MoMask) and task 4 (docs/changelog for tasks 1-3 together).
+ - Task 1: added a second generation backend, MoMask (EricGuo5513/momask-codes, MIT, confirmed via a direct LICENSE fetch), beside the existing MDM default in `workers/model-text2motion`, selectable per request via `model: "mdm" | "momask"`. New files: `motion_decode.py` (shared pure-NumPy decode math, factored out of `mdm_sampler.py` so both backends converge on the same `(T,22,3)` to local-quaternion path without either project's conflicting top-level `utils` package shadowing the other), `momask_sampler.py` (purges `sys.modules['utils'*]` and prepends `/opt/momask` to `sys.path` at load time, lazy-loaded on first `momask` request, never warmed at startup), `test_model_selector.py` (12 cases: `canonical_model()` resolution and fallback, `/infer` routing to each backend via a `FakeSampler`/`RecordingBucket` test-double pair matching `workers/model-hunyuan3d`'s own test convention, `/health` and `/` reporting both backends, the 422-vs-401 auth split). Edited: `main.py` (model selector, per-backend `motion-clips/<model>/` GCS path, `/health` and `/` now report `models_loaded`/`models_available`/`default_model`/`load_errors`), `mdm_sampler.py` (now imports the shared decode module instead of owning its own copy), `Dockerfile` (clones momask-codes, installs its one missing pip dependency `einops`, confirmed via walking every transitive import `momask_sampler.py`'s `_load()` reaches against momask-codes' own source; its `requirements.txt` additionally lists chumpy/matplotlib/scikit-learn/trimesh/gdown/vector-quantize-pytorch, none reachable from the inference path this worker calls), `cloudbuild.yaml` (adds `MOMASK_MODEL_DIR=/weights/momask` to both `deploy-primary` and `deploy-standby`), `README.md` (both backends documented end to end: contract, pipeline, weights layout, tests, build/deploy verification). Real MoMask checkpoints (188MB, downloaded via `gdown` from the project's own published Google Drive release) are staged to `gs://three-ws-model-weights/momask/t2m/` (write access probed first, probe file deleted after), matching momask-codes' own HumanML3D checkpoint directory layout.
+ - Task 4: added two `data/changelog.json` entries (newest-first, at the top of the array): one for this session's MoMask addition, one for the already-committed trim/blend/loop/save-to-library work, confirmed via grep that neither had a prior entry. Ran `npm run build:pages`; it regenerated `CHANGELOG.md`, `public/changelog.json`, `public/changelog-recent.json`, `public/changelog.xml`, `public/features.json` cleanly (`public/sitemap/index.html`, shown modified in `git status` at the start of this pass, matched the fresh i18n-annotate output byte for byte and needed no further change, meaning that pending diff was leftover build output from a prior run, not drift this pass caused). Updated the `STRUCTURE.md` fleet row for `model-text2motion` to name both backends and the `model` selector field.
+ - DoD line (a) (a prompt produces a clip previewed on a non-default avatar and saved to the library): satisfied by the already-committed task 2/3 work (`58bd13cb9`); not re-verified against a live browser this pass since no UI code changed.
+ - DoD line (b): `workers/model-text2motion`'s own suite: 49/49 passing (`python3 -m unittest discover -p 'test_*.py'` from the worker directory). Full-repo `npm test`: 2330 passed / 39 failed test files, all 39 pre-existing and unrelated (OAuth consent, MCP catalogs, grok-bot docs, `data/pages.json` duplicate-path entries from `/grok`/`/docs/grok-bot`, branding-allowlist gaps in docs this order never touched); none intersect `workers/model-text2motion`, `STRUCTURE.md`, or `data/changelog.json`. `npm run audit:docs`: 19 findings, the same pre-existing set this campaign has logged since order 067, none naming text2motion, MoMask, or this order.
+ - DoD line (c): README and STRUCTURE.md row done (above); `data/changelog.json` entries added and `npm run build:pages` run (above).
+ - `npm run check:rules -- --paths <all touched/new files>` exit 0 across every file touched this pass (30 pre-existing and newly-written em-dash/en-dash violations found and fixed across 8 files on the first two passes, 0 on the third and final pass).
+ - Owner action remaining, logged as row 46 of `production-100-OWNER-ACTIONS.md`: `gcloud builds submit`/`gcloud run deploy` are hard-denied tools in this unattended session, so the real on-GPU build, deploy, and warm-vs-warm MDM-vs-MoMask benchmark cannot run here. Code, tests, weights, and cloudbuild config are all ready; only the build+deploy+benchmark step remains. The order file stays in place (not deleted) until that step lands and the order's Definition of done line (a) is reconfirmed end to end.
+ - Commits: (this entry ships with the commits covering `workers/model-text2motion/*`, `STRUCTURE.md`, `data/changelog.json`, `public/*`, `CHANGELOG.md`, and `production-100-OWNER-ACTIONS.md`; see the commit log for this date).
diff --git a/public/changelog-recent.json b/public/changelog-recent.json
index 165425b1c..79ce95cad 100644
--- a/public/changelog-recent.json
+++ b/public/changelog-recent.json
@@ -1,11 +1,11 @@
{
- "generated_at": "2026-10-12T05:36:16.742Z",
+ "generated_at": "2026-10-12T08:21:28.589Z",
"generated_by": "scripts/build-page-index.mjs from data/pages.json + data/changelog.json",
"site": {
"name": "three.ws",
"url": "https://three.ws"
},
- "total": 4040,
+ "total": 4042,
"complete": false,
"full_feed": "/changelog.json",
"entries": [
@@ -49,6 +49,17 @@
"feature"
]
},
+ {
+ "date": "2026-10-12",
+ "type": "update",
+ "title": "Animation Studio: trim, loop-blend, and save any generated clip to your library",
+ "summary": "The Animation Studio's prompt-to-clip flow (type a prompt, generate motion on the self-hosted GPU worker, preview it retargeted onto the loaded avatar) now has an editing and persistence half. The transport grows trim start/end inputs, a loop-blend duration field, and a loop toggle, all recomputed through a non-destructive edit path so the live preview, the GLB export, and the saved clip always agree on the current edit. A Save to library button collects a name and visibility and posts the canonical, pre-retarget clip to the shared clip library, so it plays back correctly on any future rig, not just the avatar it was previewed on. A separate bug that crashed every generated clip's playback the moment it reached the browser (AnimationClip.parse() expected userData as a JSON string, the server was sending an object) is also fixed.",
+ "link": "/pose",
+ "tags": [
+ "feature",
+ "fix"
+ ]
+ },
{
"date": "2026-10-12",
"type": "update",
@@ -70,6 +81,17 @@
"fix"
]
},
+ {
+ "date": "2026-10-12",
+ "type": "update",
+ "title": "Text-to-motion gets a second GPU backend: MoMask, selectable per request",
+ "summary": "The self-hosted text-to-motion worker behind the Animation Studio and POST /api/forge-motion now runs two independent generation backends on the same GPU lane: MDM (Motion Diffusion Model, MIT), the existing default, and MoMask (EricGuo5513/momask-codes, MIT), a residual-VQ tokenizer plus masked and residual transformer that publishes a lower HumanML3D FID (0.045 vs. T2M-GPT's 0.141). Pass model: \"momask\" to /infer or /api/forge-motion to route a prompt through it instead; omit it and MDM keeps answering as before. Both backends converge on the same (T,22,3) HumanML3D position format and the same shared decode math, so a MoMask clip previews, retargets, and exports exactly like an MDM one. The two projects ship conflicting top-level `utils` packages; the worker resolves that at load time so either backend can be warmed in the same container without the other's modules leaking in. MoMask's checkpoints are staged to production storage and the worker is wired to load them; the first on-GPU deploy and a real in-house benchmark against MDM are tracked as the next step.",
+ "link": "/pose",
+ "tags": [
+ "feature",
+ "infra"
+ ]
+ },
{
"date": "2026-10-12",
"type": "update",
@@ -1523,26 +1545,6 @@
"tags": [
"launch"
]
- },
- {
- "date": "2026-10-09",
- "type": "launch",
- "title": "three.ws Forge for Desktop",
- "summary": "Download three.ws Forge, the free desktop 3D studio for Windows, Linux and Apple silicon Macs. Turn a photo into a 3D model with open-source generators on your own GPU or the three.ws cloud, chain repair, remeshing, rigging and export in node workflows, import free CC0 models and publish to your three.ws account in one click. Based on Modly (MIT).",
- "link": "/forge-desktop",
- "tags": [
- "launch"
- ]
- },
- {
- "date": "2026-10-09",
- "type": "launch",
- "title": "Tutorial · Grok Bot recipes",
- "summary": "Six jobs to hand Grok Bot, each with the exact sentence, the tools it calls in order, and the real output of running that sequence against production: a daily 3D brief, a game jam asset pack, an avatar from a photo, an X post image as a 3D model, a weekly agent report and a $THREE market brief.",
- "link": "/tutorials/grok-bot-recipes",
- "tags": [
- "launch"
- ]
}
]
}
diff --git a/public/changelog.json b/public/changelog.json
index 9ff1d1c62..e0c9d175d 100644
--- a/public/changelog.json
+++ b/public/changelog.json
@@ -1,5 +1,5 @@
{
- "generated_at": "2026-10-12T05:36:16.720Z",
+ "generated_at": "2026-10-12T08:21:28.566Z",
"generated_by": "scripts/build-page-index.mjs from data/pages.json + data/changelog.json",
"site": {
"name": "three.ws",
@@ -46,6 +46,17 @@
"feature"
]
},
+ {
+ "date": "2026-10-12",
+ "type": "update",
+ "title": "Animation Studio: trim, loop-blend, and save any generated clip to your library",
+ "summary": "The Animation Studio's prompt-to-clip flow (type a prompt, generate motion on the self-hosted GPU worker, preview it retargeted onto the loaded avatar) now has an editing and persistence half. The transport grows trim start/end inputs, a loop-blend duration field, and a loop toggle, all recomputed through a non-destructive edit path so the live preview, the GLB export, and the saved clip always agree on the current edit. A Save to library button collects a name and visibility and posts the canonical, pre-retarget clip to the shared clip library, so it plays back correctly on any future rig, not just the avatar it was previewed on. A separate bug that crashed every generated clip's playback the moment it reached the browser (AnimationClip.parse() expected userData as a JSON string, the server was sending an object) is also fixed.",
+ "link": "/pose",
+ "tags": [
+ "feature",
+ "fix"
+ ]
+ },
{
"date": "2026-10-12",
"type": "update",
@@ -67,6 +78,17 @@
"fix"
]
},
+ {
+ "date": "2026-10-12",
+ "type": "update",
+ "title": "Text-to-motion gets a second GPU backend: MoMask, selectable per request",
+ "summary": "The self-hosted text-to-motion worker behind the Animation Studio and POST /api/forge-motion now runs two independent generation backends on the same GPU lane: MDM (Motion Diffusion Model, MIT), the existing default, and MoMask (EricGuo5513/momask-codes, MIT), a residual-VQ tokenizer plus masked and residual transformer that publishes a lower HumanML3D FID (0.045 vs. T2M-GPT's 0.141). Pass model: \"momask\" to /infer or /api/forge-motion to route a prompt through it instead; omit it and MDM keeps answering as before. Both backends converge on the same (T,22,3) HumanML3D position format and the same shared decode math, so a MoMask clip previews, retargets, and exports exactly like an MDM one. The two projects ship conflicting top-level `utils` packages; the worker resolves that at load time so either backend can be warmed in the same container without the other's modules leaking in. MoMask's checkpoints are staged to production storage and the worker is wired to load them; the first on-GPU deploy and a real in-house benchmark against MDM are tracked as the next step.",
+ "link": "/pose",
+ "tags": [
+ "feature",
+ "infra"
+ ]
+ },
{
"date": "2026-10-12",
"type": "update",
diff --git a/public/changelog.xml b/public/changelog.xml
index 433d7f0a5..8584436fa 100644
--- a/public/changelog.xml
+++ b/public/changelog.xml
@@ -38,6 +38,14 @@
<category>feature</category>
<description>&lt;agent-3d&gt; now matches the AR/viewer contract the market expects from model-viewer (Apache-2.0): ar-modes picks which launch paths to try (WebXR, Scene Viewer, Quick Look), ar-placement anchors floor or wall, ios-src supplies a pre-baked USDZ or the element generates a best-effort one on device, variant-name switches a GLB's real KHR_materials_variants materials live, and any element with slot=&quot;hotspot&quot; tracks a 3D position and flies the camera to a saved orbit on click. /avatars/:id/ar and /agents/:id/ar (pages/ar.html) now run entirely on &lt;agent-3d ar&gt;; the page no longer loads a third-party viewer from a CDN.</description>
</item>
+ <item>
+ <title>Animation Studio: trim, loop-blend, and save any generated clip to your library</title>
+ <link>https://three.ws/pose</link>
+ <guid isPermaLink="false">2026-10-12:Animation Studio: trim, loop-blend, and save any generated clip to your library</guid>
+ <pubDate>Mon, 12 Oct 2026 12:00:00 GMT</pubDate>
+ <category>feature</category>
+ <description>The Animation Studio's prompt-to-clip flow (type a prompt, generate motion on the self-hosted GPU worker, preview it retargeted onto the loaded avatar) now has an editing and persistence half. The transport grows trim start/end inputs, a loop-blend duration field, and a loop toggle, all recomputed through a non-destructive edit path so the live preview, the GLB export, and the saved clip always agree on the current edit. A Save to library button collects a name and visibility and posts the canonical, pre-retarget clip to the shared clip library, so it plays back correctly on any future rig, not just the avatar it was previewed on. A separate bug that crashed every generated clip's playback the moment it reached the browser (AnimationClip.parse() expected userData as a JSON string, the server was sending an object) is also fixed.</description>
+ </item>
<item>
<title>Engine QA: a free real-time-engine fitness grade, plus real LOD chains</title>
<link>https://three.ws/docs/engine-qa</link>
@@ -54,6 +62,14 @@
<category>feature</category>
<description>The hand-rolled VRM loader is gone. vrm-boot.js now registers the real VRMLoaderPlugin (@pixiv/three-vrm) on the three.ws GLTFLoader, so VRM 0.x and 1.0 avatars get genuine MToon materials, spring-bone physics, node constraints, look-at, and all 18 expression presets, driven live by vrm.update(delta) in AnimationManager and mapped onto the ARKit-52 face-capture driver. VRMAnimationLoaderPlugin loads .vrma clips into the same retargeting pipeline that drives non-VRM humanoids, so a single clip plays on either. The Ready Player Me import page (/import-rpm) now accepts .vrm files alongside .glb, and a long-standing bug that 400'd every import at the final save step (it sent an invalid source value) is fixed. docs/avatar-engines.md upgrades three-vrm from interop to live.</description>
</item>
+ <item>
+ <title>Text-to-motion gets a second GPU backend: MoMask, selectable per request</title>
+ <link>https://three.ws/pose</link>
+ <guid isPermaLink="false">2026-10-12:Text-to-motion gets a second GPU backend: MoMask, selectable per request</guid>
+ <pubDate>Mon, 12 Oct 2026 12:00:00 GMT</pubDate>
+ <category>feature</category>
+ <description>The self-hosted text-to-motion worker behind the Animation Studio and POST /api/forge-motion now runs two independent generation backends on the same GPU lane: MDM (Motion Diffusion Model, MIT), the existing default, and MoMask (EricGuo5513/momask-codes, MIT), a residual-VQ tokenizer plus masked and residual transformer that publishes a lower HumanML3D FID (0.045 vs. T2M-GPT's 0.141). Pass model: &quot;momask&quot; to /infer or /api/forge-motion to route a prompt through it instead; omit it and MDM keeps answering as before. Both backends converge on the same (T,22,3) HumanML3D position format and the same shared decode math, so a MoMask clip previews, retargets, and exports exactly like an MDM one. The two projects ship conflicting top-level `utils` packages; the worker resolves that at load time so either backend can be warmed in the same container without the other's modules leaking in. MoMask's checkpoints are staged to production storage and the worker is wired to load them; the first on-GPU deploy and a real in-house benchmark against MDM are tracked as the next step.</description>
+ </item>
<item>
<title>Three new engine bridges: Unity, Godot, and Unreal, one prompt to a mesh in your editor</title>
<link>https://three.ws/docs/engine-bridges</link>
diff --git a/public/features.json b/public/features.json
index 54edd6565..8c6137631 100644
--- a/public/features.json
+++ b/public/features.json
@@ -1,5 +1,5 @@
{
- "generated_at": "2026-10-12T05:36:16.707Z",
+ "generated_at": "2026-10-12T08:21:28.552Z",
"generated_by": "scripts/build-page-index.mjs from data/pages.json",
"site": {
"name": "three.ws",