Turn a handful of plain photos of an object or scene into 3D Gaussians — no camera poses, no GPU required, no Python. A small C/C++ program (built on ggml) that runs the FreeSplatter neural network on your CPU (or a Vulkan GPU).
You give it N images; it gives you, for every pixel, a 3D Gaussian (position, colour, opacity, size, orientation) that a Gaussian-splatting viewer can render.
Status: under construction. The model runs are being brought up and validated one layer at a time; see
CLAUDE.mdfor the engineering approach.
nix develop # gets cmake, a compiler, and (optionally) Vulkan
cmake --preset release
cmake --build --preset releaseNo Nix? You just need CMake, a C++17 compiler, and the bundled ggml submodule
(git submodule update --init).
free_splatter-cli model.gguf images.f32model.gguf— the model weights (see Releases).images.f32— your input views as raw 32-bit floats in[0,1], laid out one view after another, each view as channels-then-rows-then-columns (RGB, 512×512). The number of views is detected automatically.
The output is one Gaussian per pixel per view. Use --out result.f32 to save it,
--device vulkan to run on a GPU.
Apache-2.0. Model weights are derived from TencentARC/FreeSplatter (Apache-2.0).