Add option for single-pass word-by-word processing to enable continous listening to commands

[ideasman42]

Add option for single-pass word-by-word processing to enable continous listening to commands

[omlins]

@ideasman42 , does this make sense to you? Would that be difficult to add?

(...and thanks for having brought this awesome project to the world!)

[ideasman42]

@ideasman42 , does this make sense to you? Would that be difficult to add?

I'm not keen on adding too much specific functionality,
there is value in keeping this small and simple allowing users to hack their own configurations to add extended functionality.

This means if users aren't happy with how a feature works (in the sample files for e.g.) they can always implemented differently in their own configuration.

As far as I can see this should not be so complicated, but it may not be a trivial task either.

---

This would be great to support, although I think this could be handled in the user configuration - with some minor extension. The examples file could include a sample which would be sufficient for most most users (so everyone doesn't have to write their own).

The way VOSK works, words may be modified as you talk (based on statistics of likely word choice).

For typing that's OK since words can be back spaced and typed in again, however it doesn't work for commands.

There would need to be a way for:

• Commands to be executed immediately, ignoring any updates from VOSK's statistical model.

  In this case the "Reset" method looks like what we want, so the statistical model is reset and previous words wont be reinterpreted.

• The user configuration to return user input such as mouse buttons (not so difficult).

[omlins]

I'm not keen on adding too much specific functionality, there is value in keeping this small and simple allowing users to hack their own configurations to add extended functionality.

Thanks for your reply! I totally share this philosophy of keeping things simple and having few, but powerful features.

I will soon come back with a solution draft in form of an update in the PR #17...