I installed PicoTTS for text to speech a long time ago, and think it sounds reasonable for a local solution. I assume that it could be used in place of Piper, or does everything have to support the Wyoming protocol for the assist pipeline to work?
I am not 100% sure of this big focus on Voice Assistants. I guess they can be good for annoucnements, but I see a voice assistant that you talk to as being nothing more than a glorified remote control, not really what a truly smart home should be. A smart home should be intelligent enough to do the stuff for you, seamlessly with well designed automations.
I agree with you in principle, but I know that I am not smart enough to anticipate every situation for myself, my family and guests. Thus I need some sort of remote control. Voice assistants have 2 really good use functions: 1 - Hands free. If I'm carrying something I can't pick my phone up to do something without putting it down. 2 - Explicit broadcast. When you say a voice command out loud everyone around you will also hear it and understand what you are doing and thus expect the comming changes. This don't replace any of my automations, remotes, dashboards... it adds to them. And another really important thing: This also gives more power to the rest of the people in your home. I'm guessing it is mostly one person setting the smart home up and building automations? Then they have all the power, and even if we do it for good we still reduce everyone elses agency in our home. This increase their agency. I don't like when people move my stuff in my home. Setting up, or changing, an automation is "moving everyones stuff". So there is a balance.
I agree. Right now all my voice stuff is pure fun. I mean. I do have a command " hey jarvis, enable barn door protocol" which arms the security system, makes sure doors are locked, closes the shades/curtains, and turns on motion notifications from any of the outside sensors. But all the main functionality is automated in this house. No dashboards. No manual triggers.
Thank you for this. Very helpful. You mentioned a couple of times that it will be possible to have different pipelines based on the Wake word. Is that possible now, or is it just outside the scope of this video?
There is a way to do it, but it's hackish and hard to set up. You need to run two instances of wyoming-satellite on each satellite device, plus two instances of wyoming-openwakeword (I run this locally on the RPi). Also, you need to make sure that you're using pulseaudio for the wyoming-satellite so that the two instances can share the same sound devices. Oh, and each instance needs to be running on a different port.
@@johnh3ss that's too bad. I wonder if it would be possible to run wake word processing on something like an RPI 2 and direct output to different pipelines from there.
@@davidhite1868 That's possible. I have an OrangePi 3 LTS right now running dual instances of both wyoming-satellite and wyoming-openwakeword. One answers to "Ok GLaDOS" and talks with a GLaDOS voice, and the other answers to "Hey Jarvis" and talks with a nice British accent. These are two separate pipelines in my Home Assistant.
Nice to hear and see you again. Just waiting to see what Nabu Casa does with a Voice smart speaker. Soon???
I installed PicoTTS for text to speech a long time ago, and think it sounds reasonable for a local solution. I assume that it could be used in place of Piper, or does everything have to support the Wyoming protocol for the assist pipeline to work?
❤👍
Will this work with HomeKit and HomePod mini?
I am not 100% sure of this big focus on Voice Assistants. I guess they can be good for annoucnements, but I see a voice assistant that you talk to as being nothing more than a glorified remote control, not really what a truly smart home should be. A smart home should be intelligent enough to do the stuff for you, seamlessly with well designed automations.
I agree with you in principle, but I know that I am not smart enough to anticipate every situation for myself, my family and guests. Thus I need some sort of remote control.
Voice assistants have 2 really good use functions:
1 - Hands free. If I'm carrying something I can't pick my phone up to do something without putting it down.
2 - Explicit broadcast. When you say a voice command out loud everyone around you will also hear it and understand what you are doing and thus expect the comming changes.
This don't replace any of my automations, remotes, dashboards... it adds to them.
And another really important thing: This also gives more power to the rest of the people in your home. I'm guessing it is mostly one person setting the smart home up and building automations? Then they have all the power, and even if we do it for good we still reduce everyone elses agency in our home. This increase their agency.
I don't like when people move my stuff in my home. Setting up, or changing, an automation is "moving everyones stuff". So there is a balance.
I agree. Right now all my voice stuff is pure fun. I mean. I do have a command " hey jarvis, enable barn door protocol" which arms the security system, makes sure doors are locked, closes the shades/curtains, and turns on motion notifications from any of the outside sensors. But all the main functionality is automated in this house. No dashboards. No manual triggers.
Thank you for this. Very helpful. You mentioned a couple of times that it will be possible to have different pipelines based on the Wake word. Is that possible now, or is it just outside the scope of this video?
Not possible now. But maybe in the future...
There is a way to do it, but it's hackish and hard to set up. You need to run two instances of wyoming-satellite on each satellite device, plus two instances of wyoming-openwakeword (I run this locally on the RPi). Also, you need to make sure that you're using pulseaudio for the wyoming-satellite so that the two instances can share the same sound devices. Oh, and each instance needs to be running on a different port.
@@johnh3ss that's too bad. I wonder if it would be possible to run wake word processing on something like an RPI 2 and direct output to different pipelines from there.
@@davidhite1868 That's possible. I have an OrangePi 3 LTS right now running dual instances of both wyoming-satellite and wyoming-openwakeword. One answers to "Ok GLaDOS" and talks with a GLaDOS voice, and the other answers to "Hey Jarvis" and talks with a nice British accent. These are two separate pipelines in my Home Assistant.