25 lines
1.4 KiB
Markdown
25 lines
1.4 KiB
Markdown
# Demo setup for Mycroft powered by Deepspeech
|
|
|
|
1. run `./setup.sh` for downloading models and scorer for deepspeech
|
|
engine
|
|
1. then `docker-compose up` is all you need
|
|
1. After Mycroft will ask you to register it on https://home.mycroft.ai/devices
|
|
|
|
## Notes
|
|
|
|
* For setup audio correctly [this](https://mycroft-ai.gitbook.io/docs/using-mycroft-ai/troubleshooting/audio-troubleshooting) was really helpfull
|
|
* Trying audio with `audiotest` is VERY helpfull `docker-compose run --entrypoint /bin/bash mycroft` then `./start-mycroft.sh audiotest -l`
|
|
* I have pulseaudio on my setup so device index 12 was pulse for me,
|
|
then I had to select the correct default device for pulse with `pactl list sources short` and `pactl set-default-source 2`. And device_index is the alsa index NOT the pulseaudio index.
|
|
* I launch the pactl commands inside the docker bash of mycroft but I
|
|
think it doesn't matter because it communicate with you host
|
|
pulseaudio server
|
|
* If you remove `stt` section in `.mycroft/mycroft.conf` Mycroft will
|
|
default to Google Speech to Text remote engine ( yes useless for
|
|
privacy )
|
|
* Deepspeech engine, is not very good, but I think some config tweak and maybe
|
|
a better stereo mic can help.
|
|
* So Mycroft only require a account on home.mycroft.ai for skills I
|
|
think, because it requires some APIs tokens for the basic skills like
|
|
weather, news, and jokes. (need to investigate on that part)
|