diff options
Diffstat (limited to 'stable-diffusion.cpp-vulkan/README')
| -rw-r--r-- | stable-diffusion.cpp-vulkan/README | 23 |
1 files changed, 23 insertions, 0 deletions
diff --git a/stable-diffusion.cpp-vulkan/README b/stable-diffusion.cpp-vulkan/README new file mode 100644 index 0000000..5f10140 --- /dev/null +++ b/stable-diffusion.cpp-vulkan/README @@ -0,0 +1,23 @@ +stable-diffusion.cpp + +Diffusion model inference in pure C/C++, built on ggml. Supports +Stable Diffusion 1.x/2.x/XL/3.x, FLUX, Wan, Qwen Image, Z-Image and +other image and video models, with LoRA, ControlNet, PhotoMaker and +ESRGAN upscaling. + +This package is built with the Vulkan backend and ships: + + sd-cli command line image/video generation + sd-server HTTP server (OpenAI and sdapi compatible) with web UI + +WebP image I/O uses the system libwebp. WebM video output is not +enabled. + +The ggml copy bundled here is upstream's patched fork, linked +statically, so it does not conflict with llama.cpp-vulkan. + +The server web UI is built offline from pinned npm tarballs listed in +the .info file; no network access is needed during the build. + +x86_64 only. Built with GGML_NATIVE, so the binaries are tuned to the +build host CPU. |