search

omegaui / linux-voice-control

Your personal, fully customization, Linux Voice Control Assistant.
GNU General Public License v3.0
132 stars 7 forks source link

Transcribing very slow and only works in live mode #14

Closed ghost closed 1 year ago

ghost commented 1 year ago

I had the same errors as in this post.

https://github.com/omegaui/linux-voice-control/issues/13#issuecomment-1682948397

I didn't do the install file as I just ran python3 main.py. It eventually started and I was in live mode and I did a test sentence: testing 1,2,3 and it transcribed it properly but it took about 5+ minutes to transcribe. I'm not sure how you could work in live mode since it picks up everything that is spoken.

I also tried python3 master_control_mode_setup.py. When I run this, it doesn't pickup any speech and eventually it will say transcribing and it takes 5+ minutes with no results.

I want to use linux voice control to control my RPi 3+ using your program. Maybe my system doesn't have enough power to run Kodi and your program. Since I need widevine, I am running in 32 bit mode and it was a dog to get pytorch, torchvision, and torchaudio compiled on a 32 bit system. I had to turn alot off like Kaldi, etc just to get it to compile and then it took about 24 hours.

omegaui commented 1 year ago

@stuckinthe Sorry for the late response, Do you have mpvlib installed? also, can you specify your system specs, also what type of storage device you are using?

ghost commented 1 year ago

Yes, I have mpvlib installed. I am running a Rpi3+ with a memory card. I just don't think the Rpi has enough to handle. Also, I'm not sure torch audio complied properly since I'm running 32 bit.

I did a test script using pocketsphhinx and it was slow as well. Vosk-api works since it has a smaller dictionary.