gk-diffuser

Image/video/audio generation GUI for GGUF diffusion models, packaged for Python — the diffusion half of the unified ggk package. The GUI runs in your browser against a local backend; each generation spawns the diffusion C/C++ CLI, compiled during pip install and bundled with the package as a single binary. The engine evaluates its graphs on gk, an independent tensor library — there is no ggml anywhere in the tree. Model files are referenced by filesystem path through a built-in file browser — nothing is uploaded or copied to temp storage.

Install

pip install gk-diffuser

Building the bundled engine requires a C/C++ toolchain and CMake ≥ 3.15 (on Windows: MSVC Build Tools). The engine source is resolved from the vendored vendor/diffusion copy (see scripts/vendor_engine.py) or GK_DIFFUSER_ENGINE_DIR. That tree is self-contained — it carries the gk kernels and the diffusion runtime with its own tokenizers and thirdparty — so the build compiles the CLI binary and nothing else.

The vendor/diffusion/gk kernels are shared verbatim with the gk-server package; both are vendored from the same unified ggk engine tree, so the two packages always compute with the same gk.

GPU and accelerator backends

The default build is CPU-only (on macOS, Metal is on by default — no flag needed). Backends are opt-in and can be requested with an environment variable or a CMake define — the env var is usually easier to pass through pip:

GK_DIFFUSER_CUDA=1   pip install gk-diffuser    # NVIDIA (needs the CUDA toolkit)
GK_DIFFUSER_HIP=1    pip install gk-diffuser    # AMD (needs ROCm/HIP)
GK_DIFFUSER_VULKAN=1 pip install gk-diffuser    # cross-vendor (needs the Vulkan SDK)

CMAKE_ARGS="-DGK_DIFFUSER_CUDA=ON" pip install gk-diffuser   # equivalent

Available: CUDA, HIP, VULKAN, METAL. Each option maps to the gk backend of the same name, and the finer-grained GK_* knobs (GK_NATIVE, GK_CUDA_ARCHITECTURES, …) can still be passed straight through as -DGK_<NAME>=….

A CUDA build works its own architecture list out from nvcc and the installed GPUs, and embeds PTX for the newest, so an unlisted card JITs rather than failing. A wheel built on one machine for another should still say what it targets, e.g. CMAKE_ARGS="-DGK_DIFFUSER_CUDA=ON -DGK_CUDA_ARCHITECTURES=89".

Run

gk-diffuser              # GUI on http://127.0.0.1:8643, opens the browser
python -m gk_diffuser    # same thing
gk-diffuser --port 0     # pick a free port; --no-browser to stay headless

The engine is directly scriptable from the CLI too:

gk-diffuser engine -- -m sd.gguf -p "a lighthouse at dusk" -o out.png
gk-diffuser engine -- --help