Turning words to numbers and avoid saving a file

[Bradams2212]

For a project, I tried using pvleopard but had to ultimately decide against it. This was because of two issues:

1. It would take the user saying 'forty four' and return the words 'forty four' instead of the number.
2. I could only use it for a limited amount of time or I would have to use speech recognition to save the file and then process the wav file.

I was wondering if there were any ways around these issues. For the second one, I would like to avoid using speech recognition but would need the program to stop listening when the user has stopped talking, not after a set amount of time. Sorry if this isn't the right place for this.

[kenarsa]

can you please elaborate on the project and what you are trying to achieve? Then I can help better.

[Bradams2212]

The project includes a voice assistant which is what I've been working on. The assistant use pvporcupine for the wake word and then takes the users voice input. This voice input is then put through multiple processes, some of which requiring numbers, not the words of numbers. It is also being run on a raspberry pi so every bit of efficiency counts. If I am able to skip the step of saving the file, I can only assume it will improve the process. So I need it to run in a way where it will stop listening to the user after they stop speaking, not after a set time limit. If it could also adjust for background noise, that would be good

[Bradams2212]

I live in Australia and I'm going to sleep now. Hopefully that information is enough

[kenarsa]

have you tried Rhino?

[Bradams2212]

How does rhino work?

[Bradams2212]

If the action is something like check the weather or get my schedule for this day, it can do that? What about more specific commands like raise a frame in tkinter

[kenarsa]

yup. Check the repo out.

[Bradams2212]

What if I wanted to turn the lights to a specific brightness or the temperature to a specific point? Does it also not limit to a timed section? Can it get a town or suburb from someones speech? In some places, I still need a speech to text conversion

[Bradams2212]

Also, the demo files only show saving the file to a wav. Is there a way around that?

[Bradams2212]

I didn't need to with porcupine. If I use a similar method will that work or is porcupine different in that respect

[Bradams2212]

I already have preexisting code to take the users voice input and interpret what they want. Would it be possible for rhino to interpret what the user wants to do, convert what they said to a string and then pass that into my pre-existing functions?

[kenarsa]

quick re-cap:

1- Leopard does not have the capability to turn written numbers into numeric form (i.e. seven to 7) and there is no immediate plan for it.
2- Rhino has that functionality. You can also mix or match it with Leopard as/if needed.
3- You should look into the header file (binding code) to learn how to use each engine. A demo is just a demo.