Google's open-source parametric 3D statistical head model with controls for identity, expression, pose and eye direction.
Updated EU status checked Last shown
Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
GNM (Generative aNthropometric Model) is a Google project aiming to provide an ecosystem of parametric 3D human models and associated perception technology. The first release, GNM Head, is a statistical 3D morphable model of the human head and face that offers disentangled control over identity, expression and head pose. It includes controllable internal anatomy: eyeballs, teeth and tongue.
According to a YouTube summary, the model exposes roughly 253 identity controls and 383 expression controls, plus head rotation and eye direction, and outputs a dense head mesh. The repository provides backends for NumPy, JAX, PyTorch and TensorFlow, along with semantic parameter sampling. A technical report is referenced in the README (arXiv 2607.23687).
The code is released under Apache 2.0, which the README describes as suitable for non-commercial and commercial use. It is aimed at researchers and developers working in computer vision, graphics and generative AI. The roadmap lists further statistical models and perception tools, but only GNM Head is available so far.
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
Parametric head model with sliders for identity (253 controls), expression (383 controls), head rotation and eye direction, outputting a dense 3D head mesh including eyes, teeth and tongue. Released under Apache 2.
Closest neighbours by meaning, not just by category.
Model
Family of four open MoE language models (19B to 280B), reportedly pruned from DeepSeek and Qwen bases.
Research · Google DeepMind
Google DeepMind research that turns a text-to-video diffusion model into a single feed-forward, multi-task vision model.
Model
Research autoencoder with a geometry-native latent space for 3D-consistent video and world generation.
Model · MiniMax
Open-weights native multimodal MoE model from MiniMax, ~428B parameters (~23B active) with 1M-token context.
Model
Research generative world model that creates consistent whole-house panorama tours from a floorplan and style reference.
Model · Google
Google's open-weights 12B multimodal (text and image input) instruction-tuned model from the Gemma 3 family.
