For who interested in using GPU for faster Whisper inference

The whisper.cpp project offers a solution for GPU-accelerated Whisper inference. Follow these steps to configure your build:

• Simply modify build_whisper.cmd:
  Locate line 55 in the build_whisper.cmd file and replace it with the following:

  cmake .. -DBUILD_SHARED_LIBS=OFF -DWHISPER_BUILD_TESTS=OFF -DWHISPER_BUILD_SERVER=ON -DGGML_CUDA=1

  This modification enables CUDA support during the build process.

• Clean Rebuild Requirement:
  If you have previously compiled whisper.cpp for CPU inference, a clean rebuild is essential for GPU integration. Simply running .\build_whisper.cmd on an existing folder will not suffice.
  Recommendation: Create a new directory, git clone meetily into this new folder, and then execute the build script. This ensures all necessary components are compiled with GPU support from scratch.

Note:

If you encounter errors related to CMake and CUDA during the build process, consider the following:

• CUDA Toolkit Installation:
  Verify that the CUDA Toolkit is correctly downloaded and installed on your system. This toolkit provides the necessary libraries and tools for CUDA development.

• Troubleshooting CMake Errors:
  If errors persist, refer to this Stack Overflow post. Copy and paste those four files to visual studio folder