Rhasspy offline voice assistant toolkit

Share your Projects! Custom Integrations
406

Is it working fine?
Were the instructions here enough or you had to do something more?

Raspberry Pi3 B (not +), using a generic USB audio dongle with a microphone attached. I plan buying a Sony PS3 Eye Camera, since it is cheap and well known for working very well.

407

Well, I must say I use it together with Snips… will post a brief description asap.

408

These are the addons I use (Rhasspy Assistant and Snowbow Wake System for Rhasspy:

image
image1534×388 50.2 KB

For microphones I use my Matrix Voice as standalone. This also outputs audio now, but that version is not yet on github.

This publishes audio over MQTT (like Snips), so I have changed my settings like this:

General:

image
image1042×1246 67.6 KB

Wake word

image
image1310×1016 95.8 KB

Audio recording:

image
image1338×802 64.9 KB

Audio playing:

image
image1470×486 45.6 KB

Config of Snowboy addon:

image
image1246×784 47.3 KB

409

I’m working on getting a new version of Rhasspy out that will hopefully address some of the outstanding issues (including pocketsphinx not responding after a few minutes). Unfortunately, illness and other things have kept me from spending as much time as I’d like on it.

My plan is to release a beta add-on that @FunkyBoT and others can try before I update the main add-on (I’m trying not to break what you have working, @Romkabouter!).

Also, @S_n_Nguy_n, I will have a Vietnamese profile for you to test soon. This new version will also add preliminary support for Greek, Hindi, and Mandarin.

410

Thanks for posting your experience. I may need to better understand your setup and maybe have a try in the Snowboy Addon again.

411

That’s bad. Keep your time. Hope everything is fine. Best wishes!

412

Don’t worry breaking anything for me, I am happy to fix it.
I am only using it for myself and no “production” environment.

Hope you feel well soon.

413

Thanks for the well wishes! I felt better today, and was able to push up most of my changes. Here’s what’s in the latest version (2.01):

  • Support for Greek, Hindi, and Mandarin
    • That makes 10 languages total (English, German, Dutch, French, Spanish, Italian, and Russian too)!
  • More snowboy settings (for the pre-trained models)
  • Parallelized portions of training process (speed up for large voice command sets)
  • Better error messages when training fails (better, not great)
  • Bug fixes related to threading, microphone usage
  • Extended Settings page with executive summary at the top

Some things that didn’t make it in this version, but are almost ready:

  • Support for MaryTTS, Flite, and PicoTTS text-to-speech engines
  • Support for Vietnamese (using Kaldi and help from Canada)
  • Option to use any profile in one of three language modes:
    1. Your voice commands only (what Rhasspy does now)
    2. Any sentence in the profile language
    3. A mixture of 1 and 2 (you weigh how much of each)

As always, let me know what I broke :slight_smile:

1 Like
414

Hi! Really really impressed by finding your project. Ordered PlayStation eye for today - really need to test and “touch” it asap)
Read your docs, and have small doubts about my Hassio (rPi 3 powered, HA external only access (duckdns&letsencrypt add-on). So my Hassio HA address is https://xxx.duckdns.org:8123
Could anybody quote proper add-on config for that case? Thanks and again. (I feel myself again in childhood getting best gift from Santa.)

415

Excellent work, will try it soon

1 Like
416

Hi, Synesthesiam
I am getting bash error (pic) wh

2019-04-08%2013_26_38-Rhasspy%20Voice%20Assistant
2019-04-08%2013_26_38-Rhasspy%20Voice%20Assistant754×67 8.36 KB
2019-04-08%2013_26_30-Rhasspy%20Voice%20Assistant en trying to update and download needed files for “ru” (made it default)

417

Hi! Nice to know you are well and back with a new version.
It appears to me that it is not stopping to work anymore. It is working for about 40min now.
I am going to buy a used Eye Camera, so I can test it a little further. Thanks!

418

Still running after 6h, so I think it’s fixed about stopping.

419

The procedure above now works?

I am having many false positives. Rhasspy captures the TV audio or music, and it triggers my automations! Even what is said has nothing to do with a saved sentence. Is there a way to increase wake word precision? And even if it is triggered, but no saved sentence is said, then it does not trigger any automation.

420

Offtopic, but my audio streamer also has audio output now!

421

On a fresh Hassio install, if I add your repository: GitHub - synesthesiam/hassio-addons: My Hass.IO add-ons

I can only see Rhasspy and not Snowboy/Mycroft.
The log gives this error:

19-04-10 20:41:28 INFO (MainThread) [hassio.addons.git] Clone add-on https://github.com/synesthesiam/hassio-addons repository
19-04-10 20:41:42 WARNING (MainThread) [hassio.addons.data] Can't read /data/addons/git/75f2ff60/mycroft-precise/config.json: required key not provided @ data['arch']. Got None
19-04-10 20:41:42 WARNING (MainThread) [hassio.addons.data] Can't read /data/addons/git/75f2ff60/snowboy/config.json: required key not provided @ data['arch']. Got None
422

I had 3 addons when included the repository. After synesthesiam answered that he removed snowboy from his files, I uninstalled snowboy add-on. I Updated HA once. Now only the Rhasspy addon appears.

423

Hmmmm, I’m not quite sure what’s happening. Do you see anything printed to the Rhasspy log inside of Hass.io? I’ll see if I can modify the code to actually return the error message.

424

Got it, thank you for the detailed feedback. I guess they decided to start requiring a list of architectures?

425

Does the Matrix Voice have a speaker?