Can i run it on pc(windows) also any one who could help would be great

[Aryanop0914]

[misutoneko]

Yes I think you're all set.
A good GPU would help of course, but your CPU is quite new and it supports AVX/AVX2.
And 16GB is sufficient even for the large model.

About Windows... well, I guess I can only offer my condolences about that.
Should be doable though, I see Windows listed as a supported platform. See this thread:
#85

Best of luck to you.

[Aryanop0914]

yeah thanks for the condolences it worked tried the base model but upgrading it large but using docker for this can you help i want a realtime transcription so i can pass it to llm generate specific answers or question which eventually drive the conversation to get specific answers .

[misutoneko]

I see. Can't really give much advice...(I barely know what docker is, lol :)
But, if you got it working for base, there's really no reason why it shouldn't work for bigger models too.
Getting realtime tho, now that's a whole other kettle of fish. I think you'll need a lot beefier setup for that.

16GB probably isn't enough to run the large model with multiple threads, either.
Anecdotally, with 16GB I generally run medium with two threads. If I use large I think I have to settle for one thread.
(Which, naturally, will be quite slow.)

[Aryanop0914]

yeah i see dont know why i jumped for large haven't seen the medium one. Also there's a catch real time means not the audio from the microphone instead my server is serving the audio in the form of mulaw chunks with rate 8000Hz lets see how can i figure it out.. trying to converting them to PCM and serve it to model..

[ytain]

You could look at this app for Windows, it is somewhat inspired by whisper.cpp
https://github.com/Const-me/Whisper