FAQ

Problem: The translation/transcript is too slow, and it shows the warning:

UserWarning: FP16 is not supported on CPU; using FP32 instead.

Answers: This means that the AI is not running on your GPU but on your CPU instead.

• The whisper AI is running best on CUDA enabled GPUs.

  NVIDIA starting from GTX 1080 should do.

• It is possible that CUDA is not installed on your System. (see Prerequisites)

Problem: The translation/transcript is still too slow. But no warning appears.

Answers:

• Your GPU might be busy with another Task or you are using a too big model for your GPU.

  Look in Task-Manager how much the GPU is used without running the AI or change to a smaller model like small or tiny using --model small or --model tiny. (see Command-line flags)

• If you happen to have a secondary PC with an GPU, you can outsource the workload to it:

  Run the AI on the secondary PC, start the Websocket-Server with --websocket_ip 0.0.0.0 to have it listen on all its IP-Addresses.

  Change the IP of the websocket-client to use to the one from the Secondary PC. (Open the html file with parameter: index.html?ws_server=ws://127.0.0.1:5000 to use secondary's PC IP-Address.)

  For OSC, give it the IP of your Primary PC using the --websocket_ip 127.0.0.1 argument and change the 127.0.0.1 to its IP.

  Stream the Audio to the Secondary PC using for example https://vb-audio.com/Voicemeeter/vban.htm

Problem: On startup, i get the following error:

RuntimeError: CUDA error: out of memory

Answers: This means that your GPU does not have enough memory to run the AI.

• Try to use a smaller model like small or tiny using --model small or --model tiny.

• If you happen to have a secondary PC with an GPU, you can outsource the workload to it. (see previous answer for more information).

Problem: On startup, i get the following error:

OSError: [Errno -9999] Unanticipated host error

Answer: This can mean that your PC is blocking access to your selected Audio device.

Or That you just have the wrong Audio Device selected. Sometimes you can find the Name of your Audio Device multiple times in the List.