I need to be fairly close to the device for it to "hear" me

[mhilbush]

I have two devices running Willow built from a repo I cloned on June 11. Each device is in a completely separate part of the house (1st floor kitchen and lower level family/rec room).

I only ever use the wake word "Alexa".

I find that I need to speak pretty loudly even when I’m relatively close to the device (5-6 feet). There's very little to no background noise in the areas where the devices are located.

Please see issue #199 for a description and photos of the environment where my devices are located.

[kristiankielhofner]

In the configuration step of the build, go to the Advanced settings menu -> Wake Word Recognition Operating Mode and select 'DET_MODE_2CH_95'. If you're feeling very adventurous you can also try 3 channel 95.

That increases the "aggressiveness" of wake word detection.

This also very likely applies to #199 (or maybe not).

By default we have extremely conservative settings and it's feedback and testing like this that helps us update our defaults to serve more users better!

[mhilbush]

Running with DET_MODE_2CH_95, this morning I had to say "Alexa" 3 times before it woke up. I was standing at the kitchen counter directly in front of the device. I'll try 3CH_95 to see if that makes a difference.

I noticed in the config that microphone gain is set to the max. I don't remember exactly what it was; something like 37 dB. I assume that's a +dB value? Is that what it should be?

[mhilbush]

I should add that after the first "Alexa", I tried to not alter the tone, volume, or pace of subsequent "Alexa"s. Of course, I'm not sure how well I actually did that since I was conscious that the first "Alexa" didn't wake it.

[kristiankielhofner]

Max gain is the default and we've done a lot of testing with it and never heard it clip or otherwise product audio quality that is problematic.

Can you make a clean video recording of this? I'd like to not only see it but also extract your audio and test/debug locally. Highest audio quality possible if you can, ideally 48kHz. If you are using wisng.tovera.io for your WIS server you can also add the URI param save_audio=true and I can grab the audio sent to WIS after wake.

[mhilbush]

Thanks. I'll do a build with save_audio=true on the URL. I'll flash both devices (I had only done 1 so far) and try to get a little more time with them. My daughter was visiting from out of town this w/e so I didn't have much time to spend testing.

[kristiankielhofner]

No worries! Much to our (very pleasant) surprise we haven't had nearly as many fundamental audio issues as we were expecting. We take them very seriously (of course) so anything you can do to help track this down would be very appreciated!

[mhilbush]

There should be a few samples to check. In one case I was about 4-5 feet away from the device. I said "Alexa, let's play pool" and it reported that it heard "Peace".

I'm curious for you to hear what the audio sounds like because I have no idea how you get "Peace." from "Let's play pool.".

Edit. Please ignore the swear words in some of the audio. I was a bit frustrated at one point lol.

[stintel]

Please try https://github.com/toverainc/willow/releases/tag/0.1.0-rc.1.