Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Activity
  • Oct 14 06:00
    Sispheor commented #643
  • Oct 14 06:00

    Sispheor on dev

    Fix gevent segmentation fault (compare)

  • Oct 14 06:00
    Sispheor closed #643
  • Oct 13 16:14
    corus87 commented #643
  • Oct 13 16:08
    Sispheor commented #643
  • Oct 13 16:08
    Sispheor commented #643
  • Oct 12 14:37
    corus87 opened #643
  • Oct 12 14:16
    corus87 commented #641
  • Oct 12 14:03
    corus87 commented #641
  • Oct 12 13:15
    Sispheor commented #641
  • Oct 12 13:08
    corus87 commented #641
  • Oct 12 13:08
    corus87 commented #641
  • Oct 12 13:03
    corus87 opened #642
  • Oct 07 19:32
    corus87 commented #641
  • Oct 07 19:11
    arnodu59115 commented #641
  • Oct 07 19:10
    arnodu59115 commented #641
  • Oct 07 11:33
    corus87 commented #641
  • Oct 07 11:23
    arnodu59115 commented #641
  • Oct 07 10:55
    corus87 commented #641
  • Oct 07 10:45
    arnodu59115 commented #641
Riseryn
@Riseryn
the model is made with daz 3d and the software with unity. it's a work in progress because it's harcoded for my network. i have a lot to do for something more versatile.
Radim Vančo
@FoxKyong
Just a question, was there some more changes needed to the current installation than just editing settings.yml?
I just edited cs and en starter kit and it works but the volume of kalliope is too low and after triggering keywords it crashes:
2020-07-31 14:53:38 :: kalliope-0.7.0 :: Say something!
Exception in thread Thread-6:
Traceback (most recent call last):
  File "/usr/lib/python3.7/threading.py", line 917, in _bootstrap_inner
    self.run()
  File "/usr/local/lib/python3.7/dist-packages/kalliope/stt/Utils.py", line 43, in run
    with self.microphone as source:
  File "/usr/local/lib/python3.7/dist-packages/kalliope/stt/SpeechRecognizer.py", line 100, in __enter__
    input=True,  # stream is an input stream
  File "/usr/local/lib/python3.7/dist-packages/pyaudio.py", line 750, in open
    stream = Stream(self, *args, **kwargs)
  File "/usr/local/lib/python3.7/dist-packages/pyaudio.py", line 441, in __init__
    self._stream = pa.open(**arguments)
OSError: [Errno -9985] Device unavailable
Radim Vančo
@FoxKyong
Never mind, I just noticed that it is problem with anything I want to play on RPi with that low sound so it looks like updating whole system did some mess with sound device.
Radim Vančo
@FoxKyong
I just pushed that cs starter kit. Can I get also write access for en starter kit or will someone else update it?
Nicolas Marcq
@Sispheor
@FoxKyong please send a PR for this one
Radim Vančo
@FoxKyong
PR sent
Radim Vančo
@FoxKyong

Ma Kalliope en 3D

Nice. Someone already did something similar here. I would also like to create something like this for Kalliope:
https://www.twitch.tv/aiamsam

Juanpa
@juanpa669
+1 It would be amazing to get it works on kalliope
Juanpa
@juanpa669
Hi all! I become crazy to try to get the right sensivity. I am currently at 0.2 and still got false detection when listen music for example. I also set it to 0.9 because I don't know if we have to increase it or decrease . I use Alexa en pdml but whatever trigger I take it's sound like the same
Patrick
@corus87
@juanpa669 use the alexa.umdl model and set sens to around 0.5 - 0.55.
Also increasing the sens means it will trigger to more sounds, decreasing means it will be harder to trigger. This means normally if you set a sens of 0.9 it will trigger everything 0.1 mostly nothing of course only if you got a good model.
don't forget to set apply_frontend to True
Juanpa
@juanpa669
Thanks @corus87 I will give a try ;)
lua86
@lua86
Snowboy has so many false positives while singing.
Patrick
@corus87
@lua86 It's not snowboy, it depends on the model and how well the model is trained
Juanpa
@juanpa669
I made short test for now, with your settings, and it seems to work great. just have to give a try with a loud background
Nicolas Marcq
@Sispheor
Somebody already played with z-wave and Somfy product with a rpi?
Riseryn
@Riseryn
Nope i use zigbee.
Juanpa
@juanpa669
no but with shelly, managed by kalliope and/or Home assistant all on rpi
Juanpa
@juanpa669
@corus87 Thanks a lot ! It works perfectly with your settings umdl etc...
Riseryn
@Riseryn
Hello, can we expect to see in a near future another self hosted stt . I would appreciate to be independant from Google or any other stt cloud based.
I also need more information on how to use the geolocation feature. Thanks.
Patrick
@corus87

@Riseryn I think the best self hosted STT option will be mozilla deepspeech, they made some big improvements lately but as far as I know deepspeech still needs a lot of cpu/gpu power to operate. Deepspeech is on my to-do list and I will keep you updated as soon as I have made progress.

But I can't tell you anything about the geolocation feature, I've never tried it.

Riseryn
@Riseryn
@corus87 Thanks for your answer. Is it possible, with the current architecture of kalliope, to have deepspeech running on another computer ? I mean like a local cloud?
Riseryn
@Riseryn
I get a quick glance to deepspeech, seems to be a real pain in the ass to have it working with other language than english and i think it will require a lot of work to have it running with Kalliope. Thanks for your investment in this project.
Patrick
@corus87
Yes it is possible to setup deepspeech on a dedicated "server" and access it with kalliope.
Once we can setup a working deepspeech server, it shouldn't be much code needed to run the stt on kalliope
Riseryn
@Riseryn
Great news :)
Riseryn
@Riseryn

I have a problem. If a sound make a false positive for the trigger, after a while and a lot of line

Google Speech Recognition could not understand audio
Say something!

I get this error

Exception in thread Thread-2006:
Traceback (most recent call last):
  File "/usr/lib/python3.7/threading.py", line 917, in _bootstrap_inner
    self.run()
  File "/home/pi/.local/lib/python3.7/site-packages/kalliope/stt/Utils.py", line 44, in run
    self.audio_stream = self.recognizer.listen(source)
  File "/home/pi/.local/lib/python3.7/site-packages/kalliope/stt/SpeechRecognizer.py", line 368, in listen
    frame_data = self._record_phrase(source, sec_per_buffer)
  File "/home/pi/.local/lib/python3.7/site-packages/kalliope/stt/SpeechRecognizer.py", line 321, in _record_phrase
    self.write_mic_level(energy, source)
  File "/home/pi/.local/lib/python3.7/site-packages/kalliope/stt/SpeechRecognizer.py", line 326, in write_mic_level
    with open(self.mic_level_file, 'w') as f:
OSError: [Errno 24] Too many open files: '/tmp/kalliope/mic_level'

Is there a way to automatically stop the trigger and come back in wait for trigger after, ie, 10 times kalliope ask "say something"?

Patrick
@corus87
It looks like it doesn't close the file properly, but actually I figured out we don't need this file, I thought it is needed to read some states of the current level but it's only used for a gui.
I will try to PR a hotfix tomorrow.
Do you know how to reproduce this particular error?
Riseryn
@Riseryn
Yes, just trigger the hotword and wait until you have the error. It may takes several minutes.
Patrick
@corus87
Thanks @Riseryn to point out this issue. I have send a PR with a fix
Riseryn
@Riseryn
@corus87 Thanks to you to fix it :)
Patrick
@corus87
you welcome :)
I also did some tests on deepspeech, english is already working quite good but other languages like german or french are like you said @Riseryn a pain in the ass, so I guess we still need to be patient
Riseryn
@Riseryn
Patience is my second name ^^
RyoKaji86
@RyoKaji86
I'm sure it'll be worth it. I'd love to be independent from Google, at least regarding everything revolving around Kalliope and home assistant ^^ keep up the great work, guys! I'm really grateful for all the dedication, effort, time and resources you put into this project. Thanks a lot!!
Riseryn
@Riseryn
I have the feeling that since the last version I have more false positive triggering the Hotword. Is it just me or anyone else have the same impression?
Patrick
@corus87

Hmm normally the false positiv depends only on your model, maybe if you have enabled apply_frontend and using a pmdl keyword which doesn't support apply_frontend. But besides that there are no big changes.

Or maybe since the new version uses pulseaudio your microphone is set to an higher volume level.

Riseryn
@Riseryn
Thanks. I will check these elements.
Riseryn
@Riseryn
I have tuned the mic in pulseaudio and it seems for the moment that there are less false positive
Riseryn
@Riseryn
I would like to be able to know the status of kalliope, ie if kalliope is awaiting the hotword or awaiting an order.
Can I get this result by the API with GET / settings / hooks ?
Patrick
@corus87
@Riseryn I don't think this is possible at the moment, but you could do something like this
hooks:
  on_waiting_for_trigger: 
    - "write-waiting-for-trigger"

  on_triggered:
    - "write-is-listening"

  - name: "write-waiting-for-trigger"
    signals: []
    neurons:
      - shell:
          cmd: "echo 'waiting_for_trigger' > trigger_state"


  - name: "write-is-listening"
    signals: []
    neurons:
      - shell:
          cmd: "echo 'is_listening' > trigger_state"
  - name: "get-trigger-state"
    signals: []
    neurons:
      - shell:
          cmd: "cat trigger_state"
          say_template: "{{ output }}"
curl -i -H "Content-Type: application/json" -X POST -d '{"mute": "True"}' http://192.168.10.100:5000/settings/mute
curl -i -H "Content-Type: application/json" -X POST http://192.168.10.100:5000/synapses/start/id/get-trigger-state
curl -i -H "Content-Type: application/json" -X POST -d '{"mute": "False"}' http://192.168.10.100:5000/settings/mute
Riseryn
@Riseryn
@corus87 Thank you very much
Juanpa
@juanpa669
Hi, I made I fresh install on raspberry pi 4...it was not a piece of fun :/
So, now I still got an error I can not find how to solve it. It is related on openweather neuron
ModuleNotFoundError: No module named 'pyowm.exceptions'
Before this error msg got module pyowm not found so I installed it
pip install pyowm
but now pyowm.exceptions module missing. I do not find any posts about this error
Nicolas Marcq
@Sispheor
pip3 install instead of pip
because pip install for python2
which is deprecated
Radim Vančo
@FoxKyong
I also use only pip which uses python3. If I need pip for both python2 and python3, then I just install pip first with python2 and then with python3 so just pip uses python3 instead of python2.
Patrick
@corus87
I can confirm that the latest version of pyowm is not working and raises the mentioned error
@juanpa669 Install version 2.10.0 as required by the neuron and it should work.
just run sudo pip3 install pyowm==2.10.0
Nicolas Marcq
@Sispheor
well seen @corus87