3D catalogue · Verified local path

G3Splat local guide

Pose-free feed-forward Gaussian splatting model for geometry reconstruction, relative pose estimation and novel-view synthesis.

Choose an app

Start with Recommended. No terminal commands are shown.

Compare 3D models
Open model filesRecommended verified starting pointOpen advanced setupPython runtime for experienced usersOpen official sourcePublisher repository and primary evidence

Official example

Play the official source preview directly below. The media stays on the official publisher server and is never hosted by LocalClaw.

What it does

Gaussian SplattingMesh ReconstructionPose EstimationNovel View Synthesis

The official implementation is tested with Python 3.10, PyTorch 2.1.2 and CUDA 11.8/12.1, with compiled Gaussian and surfel rasterizers. The released MASt3R checkpoints target 256px RealEstate10K inference; because no small-GPU table is published, 32 GB RAM and 16 GB NVIDIA VRAM are the conservative local floor.

Hardware figures are practical entry floors, not performance guarantees. Resolution, duration, precision, offloading and runtime versions can materially change memory use.

Strengths

  • MIT licensed checkpoints
  • Pose-free inference from image sets
  • 3DGS and 2DGS checkpoint variants

Limits to know

  • Scene reconstruction rather than single-object textured asset generation
  • CUDA research stack with custom rasterizers
  • VGGT-backbone release is still pending