#8667 CUDA GPU acceleration unavailable. Verify NVIDIA GPU & CUDA installation (Help -> CUDA Help).

CUDA acceleration only supports NVIDIA discrete GPUs, and it requires the driver, CUDA 12.8, and cuDNN 9.x to all be in place at the same time. Any one of these missing means the CUDA acceleration option cannot be enabled.

Troubleshooting steps:

1. Confirm your hardware: run nvidia-smi in Command Prompt. If it reports "not recognized" or no NVIDIA device is listed, it means the computer has no NVIDIA discrete GPU (AMD/Intel integrated graphics do not support CUDA) and CUDA acceleration cannot be used.
2. Update the NVIDIA driver: go to the NVIDIA official website and install the latest driver for your graphics card model. CUDA 12.8 has a minimum driver version requirement. Restart your computer after installation.
3. Install CUDA Toolkit 12.8, then install cuDNN 9.x. Copy the bin and lib directories from cuDNN into the CUDA installation directory, and add them to the system Path environment variable.
4. Restart the sp.exe in the D:/pyvideotrans directory. First verify on the software's own Help -> CUDA Help page that CUDA is available, then check CUDA Acceleration on the main interface.
5. If the CUDA check still fails after all of the above, run the Help -> CUDA Help check again and provide the prompt it outputs.

Note: If the computer only has integrated graphics or an AMD graphics card, this prompt is normal. Just uncheck CUDA Acceleration and run on CPU. In that case, switch the speech recognition model to medium/small, or reduce Beam Size and Best Of to 1, to shorten processing time.

Please refer to the related documentation:

• Configuring CUDA and cuDNN on Windows