Welcome to whisper_streaming_web Discussions!

[QuentinFuxa]

👋 Welcome!

We’re using Discussions as a place to connect with other members of our community. We hope that you:

• Ask questions you’re wondering about.
• Share ideas.
• Engage with other community members.
• Welcome others and are open-minded. Remember that this is a community we
  build together 💪.

To get started, comment below with an introduction of yourself and tell us about what you do with this community.

[Farbdose]

@QuentinFuxa

Hello folks, im I'm Farbdose (severly) handycapped atm and use voice recognition to control my pc most of the time.

Because I majored im computer science and am generally a hands on guy I wrote myself a custom bridge between google voice API and xdotool to perform actions on my pc... Now I'm running into the problem that this is getting kind of expensive and was looking for alternatives to run the recognition locally on my Nvidia card.

Im Ininterested in using Whisper life Kit for my own use and am willing to write pull requests if necessary to get the functionality I need

[QuentinFuxa]

Hi @Farbdose,
Of course!
How do you usually proceed?
Would you like to control both the keyboard and the mouse movements with it?

[Farbdose]

Hi @QuentinFuxa,
sorry for the late reply. I kind of forgott about my post after a week.
I currently use Google Voice API for fast voice recognition because it provides really fast first interpretations of what is being said. For maus actions I listen to the first word being click or the like and execute pre definded rules immediately. Otherwise I wait for the API to tell me that the speaker had a little break of about one to two seconds.
Actions range from little keyboard shortcuts to whole scripts basically everything I can a write in Python.

My main issue is that I require a stream of words being said, in chunks that provide isFinal information.

which is kind of difficult with the current WhisperLiveKit with the engine always sending the whole state.
Besides, sometimes parts that were already said get updated retroactively.

I don't need to know what happened 10 seconds ago, but I need accurate information if the person who is speaking stopped speaking.

I tried to ask here #288 but got no response unfortunately.