Add openvino support for frigate 0.17 audio transcription

[vladtaylor]

Currently the audio transcription runs on either CPU or CUDA, please add VAAPI support for discreet intel GPU's

[NickM-27]

I'm assuming you mean openvino, this will likely have to come in a future version as we'll need to get the models in a different format

[vladtaylor]

sorry yes faster-whisper with OpenVINO , Having some mixed results but this drops cpu usage massively.

Would also like to get speaker-diarization working and sound characterisation ie (beeps, sqeaks, bangs etc)