How to keep model in high alert state for faster inference?

[SuperMaximus1984]

I'm running insanely-fast-whisper in environment where low latency is crucial. Once .wav file is created, it needs to be transcribed immediately. Everytime I run:

D:\InsanelyFastWhisper>insanely-fast-whisper --file-name myoutputfile_17245952653c006.wav --device-id 0
Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained.
🤗 Transcribing... ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 0:00:00
Voila!✨ Your file has been transcribed go check it out over here 👉 output.json

I see some pause in this part: Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained.

And then Transcribing runs pretty fast. I assume the model is being loaded first, so it takes time.
How do I keep it in the high alert state so that transcription would run faster upon feeding of the new .wav file?
Thanks!

[ArmykOliva]

Don't use the CLI and use the transformers python way.