diff --git a/CMakeLists.txt b/CMakeLists.txt
index ffee1ccc..8b4264fa 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -10,6 +10,58 @@ if (GGML_CUDA OR GGML_METAL)
     add_custom_target(create_empty_file ALL DEPENDS ${EMPTY_FILE_PATH})
 endif()
 
+# Project: stable_diffusion_cpp
+project(stable_diffusion_cpp)
+
+option(STABLE_DIFFUSION_BUILD "Build stable-diffusion.cpp shared library and install alongside python package" ON)
+
+if (STABLE_DIFFUSION_BUILD)
+    set(BUILD_SHARED_LIBS "ON")
+    option(SD_BUILD_SHARED_LIBS "" "ON")
+
+    # Building llama
+    if (APPLE AND NOT CMAKE_SYSTEM_PROCESSOR MATCHES "arm64")
+        # Need to disable these llama.cpp flags on Apple x86_64,
+        # otherwise users may encounter invalid instruction errors
+        set(GGML_AVX "Off" CACHE BOOL "ggml: enable AVX" FORCE)
+        set(GGML_AVX2 "Off" CACHE BOOL "ggml: enable AVX2" FORCE)
+        set(GGML_FMA "Off" CACHE BOOL "ggml: enable FMA" FORCE)
+        set(GGML_F16C "Off" CACHE BOOL "ggml: enable F16C" FORCE)
+    endif()
+
+    add_subdirectory(dependency/stable-diffusion.cpp)
+    install(
+        TARGETS stable-diffusion
+        LIBRARY DESTINATION ${SKBUILD_PLATLIB_DIR}/nexa/gguf/lib
+        RUNTIME DESTINATION ${SKBUILD_PLATLIB_DIR}/nexa/gguf/lib
+        ARCHIVE DESTINATION ${SKBUILD_PLATLIB_DIR}/nexa/gguf/lib
+        FRAMEWORK DESTINATION ${SKBUILD_PLATLIB_DIR}/nexa/gguf/lib
+        RESOURCE DESTINATION ${SKBUILD_PLATLIB_DIR}/nexa/gguf/lib
+    )
+
+    message(STATUS "SKBUILD_PLATLIB_DIR: ${SKBUILD_PLATLIB_DIR}")
+    # Temporary fix for https://github.com/scikit-build/scikit-build-core/issues/374
+    install(
+        TARGETS stable-diffusion
+        LIBRARY DESTINATION ${CMAKE_CURRENT_SOURCE_DIR}/nexa/gguf/lib
+        RUNTIME DESTINATION ${CMAKE_CURRENT_SOURCE_DIR}/nexa/gguf/lib
+        ARCHIVE DESTINATION ${CMAKE_CURRENT_SOURCE_DIR}/nexa/gguf/lib
+        FRAMEWORK DESTINATION ${CMAKE_CURRENT_SOURCE_DIR}/nexa/gguf/lib
+        RESOURCE DESTINATION ${CMAKE_CURRENT_SOURCE_DIR}/nexa/gguf/lib
+    )
+    # Workaround for Windows + CUDA
+    if (WIN32)
+        install(
+            FILES $<TARGET_RUNTIME_DLLS:stable-diffusion>
+            DESTINATION ${SKBUILD_PLATLIB_DIR}/nexa/gguf/lib
+        )
+        install(
+            FILES $<TARGET_RUNTIME_DLLS:stable-diffusion>
+            DESTINATION ${CMAKE_CURRENT_SOURCE_DIR}/nexa/gguf/lib
+        )
+    endif()
+endif()
+
 # Project: llama_cpp
 project(llama_cpp)
 
@@ -132,3 +184,4 @@ if (LLAMA_BUILD)
         endif()
     endif()
 endif()
+