#8439 CUDA GPU acceleration unavailable. Verify NVIDIA GPU & CUDA installation (Help -> CUDA Help).

CUDA acceleration is unavailable because the software did not detect a usable NVIDIA + CUDA 12.8 + cuDNN 9.11 environment. Check the following three items in order:

1. Confirm the hardware is an NVIDIA GPU (N-card). AMD, Intel integrated graphics, and Intel discrete GPUs cannot use CUDA at all. In Command Prompt run nvidia-smi: if this command errors or is not found, your driver/GPU environment is not usable.
2. Update the NVIDIA driver from the NVIDIA official website to the latest version. CUDA 12.8 has a minimum driver version requirement; an old driver reports "unavailable" even when the toolkit is installed.
3. Install cuDNN 9.x (9.11). CUDA Toolkit 12.8 alone is not enough — the cuDNN bin and lib directories must be fully added to the system Path environment variable, otherwise the software still cannot start GPU inference. After installing, restart the software.

Note: on a laptop with both integrated graphics and a discrete NVIDIA GPU, the software uses the GPU with index 0 by default; if that device is the weak integrated GPU, CUDA detection also fails. In Windows graphics settings, set the high-performance NVIDIA GPU as the preferred device for sp.exe.

If CUDA still cannot be enabled, run on CPU with a smaller model: set STT Service to faster-whisper(Built-in), Model to medium/small, and set Compute Type in Advanced Options - Speech Recognition (ASR) to int8.

Additional note (unrelated to CUDA but a risk): your software root directory is C:/Users/P/OneDrive/Pictures, i.e. a OneDrive-synced personal folder. Move the software to a shallow local path containing only English letters and digits (e.g. D:/pyVideoTrans), otherwise OneDrive syncing and path restrictions will cause failures at the video output and temp file stages.

Please refer to the related documentation:

• Configuring CUDA and cuDNN on Windows
• faster-whisper (Local, Built-in)
• FAQ