If DRE speech recognition uses your CPU to transcribe speech to text, here are some common tips to resolve this and start utilizing your GPU:
- Ensure you have the latest Microsoft Visual C++ Redistributable (more info)
- Ensure for CUDA inference to have either
- the cudart64_12.dll file in DRE’s app data (check in
%LocalAppData%/DRE) - installed a CUDA runtime between 12.1 – 12.9 (make sure to restart your PC)
- the cudart64_12.dll file in DRE’s app data (check in
- Vulkan inference requires no custom installation, so consider switching to this:
- In DRE, head to `Sound -> Recognition -> Preferred Neural Hardware Priority`
- Change the order from CUDA, Vulkan, CPU -> Vulkan, CUDA, CPU to prefer Vulkan over CUDA