Build an open future with us.

Invest in Mycroft and become a community partner.

Mycroft switches between 2 voices


#1

i have my mycroft set to use the premium voice. When i ask it to tell me a joke, it uses this voice… However if i ask it what time it is, then it uses the default male english voice. ( wich sounds terrible by the way)

when i ask it what the weather is like, i get the default male voice, but when i ask it to turn of my lights, it uses the premium voice…

So does my mycroft suffer from a split personality disorder or this this a bug? :smile:


#2

the old voice will be used in case of network errors, if the connection fails it will fallback to mimic1

if it only happens in specific sentences then it might be a good idea to open an issue, it might be some error in the backend


#3

Definitely check the logs and see what’s going on there.


#4

I did a fresh install of the latest Picroft image today and I have just encountered the same issue. It switched to default male when I asked for the weather but then switched back to premium male when I asked it to recite something. I then tried alternating between the ‘say something’ and the ‘configuration update’ commands and it wasn’t consistent with its voice choice between these either. The same issue occurs if the female voice was initially chosen.


#5

Can you get some logs on this?


#6

Today both mark_1 and picroft in my house has Allan Pope voices when asking what time, but asking for configuration update it is the Premium voice.
It feels like cached phrases is OK, but new ones isnt. The time used indicates that it is trying to get the TTS from mycroft.ai.

What to look for in log(and which log)?


#7

in audio.log i see this when asking “what time is it”

12:13:45.259 - urllib3.connectionpool - DEBUG - Starting new HTTPS connection (1): api.mycroft.ai:443
Playing WAVE '/tmp/mycroft/cache/tts/e8a7707c0da953c76f74b88454865730.wav' : Signed 16 bit Little Endian, Rate 44100 Hz, Mono
12:13:45.756 - urllib3.connectionpool - DEBUG - https://api.mycroft.ai:443 "POST /v1/device/1bbbafdd-e5c6-46ad-a0ae-732780bf9185/metric/timing HTTP/1.1" 200 279
12:13:48.089 - urllib3.connectionpool - DEBUG - Starting new HTTPS connection (1): api.mycroft.ai:443
12:13:48.552 - urllib3.connectionpool - DEBUG - https://api.mycroft.ai:443 "POST /v1/device/1bbbafdd-e5c6-46ad-a0ae-732780bf9185/metric/timing HTTP/1.1" 200 224
12:14:34.993 - mycroft.audio.speech:mute_and_speak:119 - INFO - Speak: It is thirteen fourteen
12:14:35.020 - urllib3.connectionpool - DEBUG - Starting new HTTPS connection (15): mimic-api.mycroft.ai:443
12:14:40.390 - mycroft.audio.speech:mute_and_speak:123 - ERROR - Mimic 2 remote server request timedout. falling back to mimic
12:14:40.412 - urllib3.connectionpool - DEBUG - Starting new HTTPS connection (1): api.mycroft.ai:443
12:14:40.879 - urllib3.connectionpool - DEBUG - https://api.mycroft.ai:443 "GET /v1/device/1bbbafdd-e5c6-46ad-a0ae-732780bf9185/subscription HTTP/1.1" 200 198
[('4', 0.2), ('0', 0.275), ('3', 0.362), ('0', 0.411), ('3', 0.502), ('3', 0.57), ('2', 0.681), ('3', 0.802), ('0', 0.937), ('3', 1.018), ('5', 1.129), ('6', 1.275), ('2', 1.292), ('3', 1.377), ('0', 1.573), ('3', 1.701), ('4', 2.0)]
12:14:41.883 - urllib3.connectionpool - DEBUG - Starting new HTTPS connection (1): api.mycroft.ai:443
Playing WAVE '/tmp/mycroft/cache/tts/dd461aba7da276ad0104187e75c210bd.wav' : Signed 16 bit Little Endian, Rate 44100 Hz, Mono
12:14:42.354 - urllib3.connectionpool - DEBUG - https://api.mycroft.ai:443 "POST /v1/device/1bbbafdd-e5c6-46ad-a0ae-732780bf9185/metric/timing HTTP/1.1" 200 272
12:14:43.949 - urllib3.connectionpool - DEBUG - Starting new HTTPS connection (1): api.mycroft.ai:443
12:14:44.436 - urllib3.connectionpool - DEBUG - https://api.mycroft.ai:443 "POST /v1/device/1bbbafdd-e5c6-46ad-a0ae-732780bf9185/metric/timing HTTP/1.1" 200 223

My gues is that it is this problem:
ERROR - Mimic 2 remote server request timedout.

I did change voice on home.mycroft.ai to se if it helped, but it didnt. now all replys is just the AP voice.


#8

@ake-mycroft is confirming (on mattermost) that there is a problem with mimic2 host. So as soon someone has gotten that fixed I am sure Mycroft will use the premium voice again.


#9

The gret people at Mycroft has fixed a problem with DNS at the backend. And now mycroft do speak with prenium voices again.

The incidents report can be located here


#10

Confirming we had an outage of the server that provides the mimic2 voices.
This outage is now resolved.

We have architected this so that if there is an internet connectivity failure, or a failure of the mimic2 server, then Mycroft will “fall back” to the local version of mimic. There is an additional layer here were common phrases are “cached” locally - this is what gave rise to some phrases being different voices.

We will look at the root cause of the outage and what we can do to prevent a recurrence.