OpenWakeWord doesn't recognize any wake word

Johnnii360 · April 14, 2025, 10:45am

Hey guys!

I try now since yesterday to bring OpenWakeWord or Assist on my Home Assistant to work. I tried to fix this issue with ChatGPT first, but it seems this AI is overwhelmed with that.

Problem

OpenWakeWord doesn’t recorgnize any wake word. Piper, Whisper and OWW is correctly installed and connected to Wyoming Protocol.

OWW only connects and disconnects like in the log shown and flooding it with it:

s6-rc: info: service s6rc-oneshot-runner: starting
s6-rc: info: service s6rc-oneshot-runner successfully started
s6-rc: info: service fix-attrs: starting
s6-rc: info: service fix-attrs successfully started
s6-rc: info: service legacy-cont-init: starting
s6-rc: info: service legacy-cont-init successfully started
s6-rc: info: service openwakeword: starting
s6-rc: info: service openwakeword successfully started
s6-rc: info: service discovery: starting
DEBUG:root:Namespace(uri='tcp://0.0.0.0:10400', models_dir=PosixPath('/usr/local/lib/python3.11/dist-packages/wyoming_openwakeword/models'), custom_model_dir=['/share/openwakeword'], preload_model=['ok_nabu'], threshold=0.5, trigger_level=1, output_dir=None, debug=True, log_format='%(levelname)s:%(name)s:%(message)s', debug_probability=False, version=False, model=[])
DEBUG:root:Loading ok_nabu_v0.1 from /usr/local/lib/python3.11/dist-packages/wyoming_openwakeword/models/ok_nabu_v0.1.tflite
DEBUG:wyoming_openwakeword.handler:Started thread for ok_nabu_v0.1
DEBUG:root:Loading /usr/local/lib/python3.11/dist-packages/wyoming_openwakeword/models/melspectrogram.tflite
DEBUG:root:Loading /usr/local/lib/python3.11/dist-packages/wyoming_openwakeword/models/embedding_model.tflite
INFO:root:Ready
DEBUG:wyoming_openwakeword.handler:Client connected: 1501232393578
DEBUG:wyoming_openwakeword.handler:Client disconnected: 1501232393578
[11:27:58] INFO: Successfully sent discovery information to Home Assistant.
s6-rc: info: service discovery successfully started
s6-rc: info: service legacy-services: starting
s6-rc: info: service legacy-services successfully started
DEBUG:wyoming_openwakeword.handler:Client connected: 1504611613639
DEBUG:wyoming_openwakeword.handler:Sent info to client: 1504611613639
DEBUG:wyoming_openwakeword.handler:Client disconnected: 1504611613639
DEBUG:wyoming_openwakeword.handler:Client connected: 1536692992465
DEBUG:wyoming_openwakeword.handler:Sent info to client: 1536692992465
DEBUG:wyoming_openwakeword.handler:Client disconnected: 1536692992465
DEBUG:wyoming_openwakeword.handler:Client connected: 1568768290671
DEBUG:wyoming_openwakeword.handler:Sent info to client: 1568768290671
DEBUG:wyoming_openwakeword.handler:Client disconnected: 1568768290671
DEBUG:wyoming_openwakeword.handler:Client connected: 1600838788126
DEBUG:wyoming_openwakeword.handler:Sent info to client: 1600838788126

The USB microphone works very well with Rhasspy and is set as default recording device in HA.

I also looked into the /share-folder and there’s no openwakeword-folder. OpenWakeWord also tells me nothing about an audio device.

System informations

OS Version: Home Assistant OS 15.1
Home Assistant Core: 2025.4.2
Device: ODROID N2+
Microphone: SPC Gear SM950T Streaming USB-Microphone

If you need more informations, please tell me.

ViggoVH · April 20, 2025, 9:36am

Same problem here.

will35 · April 20, 2025, 10:44am

Hi
Is Assist-Microphone Addon installed ?

Johnnii360 · April 20, 2025, 12:34pm

Assist-Microphone doesn’t use the Wyoming Protocol and doesn’t use wake words either.

will35 · April 20, 2025, 2:05pm

Assist-Microphone is recognized by Wyoming Protocol and uses the pipeline you defined. It’s been working for me for a year with usb conference microphone.

Johnnii360 · April 20, 2025, 3:09pm

Hmm… Than ChatGPT lied. But I installed it in the past but couldn’t get it to run because there is no port to integrate it into Wyoming Protocol. Or is there another way? There’s also nothing written in the documentation about this. Wyoming Protocol doesn’t detect it either.

will35 · April 20, 2025, 4:11pm

Just launch the addon in debug mode to find the port . No need ChatGPT for that

uri='tcp://0.0.0.0:10700', name='assist microphone'

Assist-Microphone port is 10700

Johnnii360 · April 20, 2025, 6:29pm

Yay it works now! It’s important to user the hostname of the container instead of localhost. It’s also necessary to set it the middle way (hard coded phrases). But maybe it also works with full local but needs a lot of performance.

Forgot to say thank you. Here you are.

will35 · April 21, 2025, 10:49am

Hello

For full local STT , if you don’t speak english , you can try Vosk with default model and conf. It is very fast and accurate ( less than 2 sec on RPI4 ) :
hassio-addons/vosk/DOCS.md at master · rhasspy/hassio-addons

and for Wakeword , use Micro-Wakeword addon with “Okay Nabu”

Johnnii360 · April 21, 2025, 6:13pm

Speech-to-Phrase (the middle way) is firstly enough for me and supersedes the deprecated Rhasspy model. For fully local STT my ODROID N2+ is maybe to slow. As written in the docs an Intel NUC is a good choice here.

But I use Piper as Text-to-Speech. Karlsson is the only male voice that I can accept. The female voices are partly good but Ramona sound like a old lady. It would be very nice if were there a voice of Jarvis from Iron Man in German and English or VEGA from Doom Eternal.