GNM-Maya 1.3.0

The exploration release: pick faces from a contact sheet, sculpt by feature zone, step back through every look, drag landmarks live, and export mocap-ready rigs — for Google GNM parametric heads inside Autodesk Maya (Windows / Linux / macOS).

What's new since 1.2.0

• Variants contact sheet — a 3×3 grid of candidate randomizations rendered as thumbnails; click a face to apply it, Re-roll for nine more. Respects the Area/Zone mask, random scale and the Symmetry toggle.
• Feature zones (Nose / Mouth / Jaw / Brows / Eyes) — randomize finer areas than the model's own regions. A fit solver blends the change into a smooth landmark-derived zone and solves back to ordinary coefficients, so sliders, presets and rig baking stay fully consistent. Isolation is soft by design: a small natural falloff halo remains (measured ~1 mm on a 4–17 mm zone change, <0.3 mm far away).
• Area Randomize masks (from the previous dev drop) — per-region checkboxes (head, eyes, lower face, …) so randomize/reset touch only what you choose.
• History ladder (◀ ▶) — step back/forward through every randomize, sample, describe, variant, reset and fit. A good face is never lost again.
• Live landmark fit (drag-to-sculpt) — GNM ▸ Landmarks ▸ Toggle Live Fit: drag the 68 locators and the head re-solves its identity on every drag release.
• ARKit-52 target names on Bake Rig — optional renaming (and Left/Right splitting of symmetric shapes with a smooth midline mask) of the semantic blendshape targets to Live Link Face conventions: eyeBlinkLeft/Right, mouthSmileLeft/Right, cheekPuff, tongueOut, … 28 targets from the 20 GNM expressions; unmapped shapes keep their GNM names.

Install (Maya 2022–2027, Windows / Linux / macOS)

1. Download Source code (zip) below and extract it anywhere.
2. Drag drag_and_drop_install.py (top of the extracted folder) into a running Maya viewport.
3. Accept the first-run setup (Python runtime + GNM model ~40 MB; shape images render locally ~5 min).

The installer copies the module into ~/Documents/maya/modules, so you can delete the zip afterwards. Updating from an earlier version: GNM ▸ Check for gnm-maya Tool Updates, then restart Maya.

Notes

• Linux/macOS need a python3 (3.9+, venv module) on PATH; Maya's own Python is the fallback. Windows is fully self-contained.
• GNM is Apache-2.0 (Google); this tool is MIT and not affiliated with Google. See NOTICE.md.

GNM-Maya 1.2.0

GNM-Maya goes cross-platform: Windows, Linux and macOS from one codebase — the full artist toolset wrapping Google GNM parametric human heads inside Autodesk Maya.

What's new since 1.1.0

• Linux & macOS support. The module's own Python runtime is now built per platform:
  • Windows — unchanged: a portable embeddable CPython is downloaded once (self-contained, relocatable, no system Python needed).
  • Linux/macOS — a venv is created inside the module on first run from your python3 (3.9+), with Maya's own mayapy as an automatic fallback. All packages stay under the module's runtime/ folder — nothing touches your system.
• The drag-and-drop installer, first-run setup, shape-gallery renderer, photo-fitting installer and dev rebuild script are all platform-aware; the installer knows a posix venv isn't relocatable and rebuilds it in place instead of copying a broken one.
• Docs updated with per-OS requirements and notes.

Install (Maya 2022–2027)

1. Download Source code (zip) below and extract it anywhere.
2. Drag drag_and_drop_install.py (top of the extracted folder) into a running Maya viewport.
3. Accept the first-run setup (Python runtime + GNM model ~40 MB; shape images render locally ~5 min).

The installer copies the module into ~/Documents/maya/modules, so you can delete the zip and extracted folder afterwards. Everything runs offline after setup.

Per-platform requirements

• Windows — nothing extra; the runtime is fully self-contained.
• Linux — python3 (3.9+) with the venv module (Debian/Ubuntu: sudo apt install python3-venv); mayapy is the fallback.
• macOS — python3 (3.9+) via Xcode CLT or Homebrew; mayapy is the fallback.

Notes

• The Windows path is fully validated (headless test suite); Linux/macOS follow the same code with platform branches unit-tested — if you hit an issue on those platforms, every failure now surfaces in an error dialog, and reports are welcome.
• Photo fitting (optional ~290 MB MediaPipe/OpenCV) was validated on Windows; the same install is attempted on Linux/macOS.
• GNM is Apache-2.0 (Google); this tool is MIT and not affiliated with Google. See NOTICE.md.

GNM-Maya 1.1.0

Second release of GNM-Maya — a full artist toolset wrapping Google GNM parametric human heads inside Autodesk Maya.

This is a UX-focused update on top of v1.0.0: the whole tool now looks and reads better, installs more obviously, and updates the way you choose.

What's new since 1.0.0

• Material-style icons across the UI — every panel button, the shelf button, the window, and all GNM menu items now carry consistent monochrome icons, tinted to Maya's dark theme. Tooltips were rewritten throughout as concise titled hints.
• Reset in the Semantic tab — Sample Identity and Sample Expression each gained a paired Reset, so you can toggle a sampled look back to neutral (and it clears the matching blend Mix too).
• Installer front-and-centre — drag_and_drop_install.py now sits at the top of the extracted download instead of buried in a subfolder, so it's the first thing you see.
• No more blank white progress bar — the first-run setup (and rig bake / crowd / photo-deps) progress window now paints its status before the work begins, instead of flashing white.
• Release / Dev update channels (from 1.0.0, now the default path) — tool updates follow tagged Releases by default; switch to the Dev channel in the update dialog to track master commits.

Install (Windows, Maya 2022–2027)

1. Download Source code (zip) below and extract it anywhere.
2. Drag drag_and_drop_install.py (top of the extracted folder) into a running Maya viewport.
3. Accept the first-run downloads (portable Python runtime ~70 MB + GNM model ~40 MB; shape images render locally ~5 min).

The installer copies the module into ~/Documents/maya/modules, so you can delete the zip and extracted folder afterwards. Everything runs offline after setup.

Updating from 1.0.0

Use GNM ▸ Check for gnm-maya Tool Updates and restart Maya, or just re-drag the installer. (If you updated between releases and icons didn't appear, this release fixes that — the icons now ship inside the code package.)

Notes

• Windows only (portable runtime is win-amd64). Requires internet for first-run setup.
• GNM is Apache-2.0 (Google); this tool is MIT and not affiliated with Google. See NOTICE.md.

GNM-Maya 1.0.0

Generate Google GNM parametric human heads inside Autodesk Maya — a full artist toolset around the GNM Head 3D Morphable Model.

Install (Windows, Maya 2022–2027)

1. Download Source code (zip) below and extract it anywhere.
2. Drag gnm-maya/drag_and_drop_install.py into a running Maya viewport.
3. Accept the first-run downloads (portable Python runtime ~70 MB + GNM model ~40 MB; shape images render locally ~5 min).

The installer copies the module into ~/Documents/maya/modules, so you can delete the zip and extracted folder afterwards. Everything runs offline after setup.

Highlights

• Live parametric head: quad mesh with UVs and per-part materials; tabbed panel with Identity (253) / Expression (383) / Pose / Translation sliders, ~15 ms interactive updates via a resident model worker
• Visual sliders: every mode shows min/max shape thumbnails (resizable), with a browsable gallery (HTML + GitHub markdown) including each mode's strength (max Δ in mm)
• Semantic controls: sample identity by gender × ethnicity, 20 named expressions, mix sliders for blending, and a natural-language Describe field (local lexicon, optional Ollama)
• Fit from Photo: 68-landmark detection (MediaPipe, fully local, optional ~290 MB install) + least-squares identity fit
• Landmark sculpting: 68 barycentric locators with L/R mirror editing and "Fit Head to Locators"
• Bake Rig: self-sufficient blendShape + skinned joint asset (semantic and/or per-region basis targets), FBX export preset
• Production extras: presets library, crowd generator, configurable folders, progress bars, self-healing on deleted meshes
• Self-updating: separate update checks for the GNM model and this tool. Tool updates follow Releases by default, with an optional Dev channel that tracks master-branch commits
• Clear failures: every error surfaces as a critical message dialog, not just a Script Editor traceback

Notes

• Windows only (portable runtime is win-amd64). Requires internet for first-run setup.
• GNM is Apache-2.0 (Google); this tool is MIT and not affiliated with Google. See NOTICE.md.