#2949 A child process terminated abruptly, the process pool is not usable anymore

This error is caused by insufficient video memory (VRAM) when running the faster-whisper model for speech recognition.

Solution:

1. Reduce VRAM usage:

   • Open the software and go to Menu -> Tools -> Advanced Options.
   • In the Speech Recognition Parameters section, find the Batch Size setting for faster-whisper and change it to 1.
   • Save the settings and restart the software.

2. Use a smaller model (if step 1 is not enough):

   • In the main interface, under Speech Recognition Channel, select faster-whisper.
   • Click the Settings button next to it.
   • Change the Model from large-v2 or large-v3 to a smaller one like medium or small. This significantly reduces VRAM requirements but may slightly lower accuracy.

3. Enable model quantization (advanced):

   • In the faster-whisper settings, you can try enabling int8 quantization. This reduces the model's memory footprint.
4. Close other GPU-intensive applications to free up VRAM.

If the problem persists after these steps, it indicates your GPU's VRAM is too limited for local speech recognition. Consider these alternatives:

• Use an online speech recognition API (e.g., OpenAI Whisper API, Qwen3-ASR from Alibaba Bailian) which offloads the processing to remote servers.
• Ensure you have the latest NVIDIA drivers installed.

Please refer to the relevant documentation:

• 语音识别模型分类与说明
• 总是提示显存不足