google/nerfies

Health v2 · maintenance only

Flags: archived

Adoption not part of the score

1973 stars · 277 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

Nerfies is the official JAX implementation of Deformable Neural Radiance Fields, which reconstructs non-rigidly deforming scenes from casually captured phone videos. It augments NeRF with a continuous deformation field and elastic regularization to produce free-viewpoint, photorealistic renderings of subjects like selfie portraits.

Use cases

• turn selfie videos into free-viewpoint 3D portraits
• reconstruct deformable scenes from casual phone captures
• train a deformable NeRF model from a video dataset
• render novel viewpoints of a non-rigid subject
• create dolly zoom and other visual effects
• extract mattes of a subject during rendering
• animate between captured states via interpolation

When to choose

• you need photorealistic novel views of a deforming subject from phone video
• you want a research-grade reference implementation of deformable NeRF in JAX
• you have GPU or TPU resources for offline training
• you want to build on or extend NeRF deformation research

When to avoid

• you need real-time rendering or interactive performance
• you want a plug-and-play app with no training setup
• you work outside Python/JAX ecosystems
• you need actively developed features or support, since updates are infrequent

Facets

library · maturity maintenance

machine-learning graphics image-processing simulation computer-vision graphics machine-learning deep-learning python cross-platform nerf neural-rendering jax 3d-reconstruction deformable-scenes novel-view-synthesis research-code gpu linux

Member repositories

For agents

markdown · JSON · MCP: product_card(name="google/nerfies")