Custom Speech-to-Text (STT) and Text-to-Speech (TTS) Servers for Mycroft AI | Digi-Key Electronics

Подписаться 159 тыс.

Просмотров 68 тыс.

50% 1

Additionally, you will need a separate computer to host your STT and TTS servers. While they might run on a Pi, they will be incredibly slow. I highly recommend using a computer with a CUDA-capable Nvidia graphics card, which will speed up the STT and TTS processes.
In the first part of the guide, we show you how to install the Mozilla DeepSpeech program, which performs speech-to-text. We run it inside a web server, which we enable on boot through systemd. From here, we can send audio data (including .wav files) to the server to have it respond with text data (in string format).
Next, we install Coqui TTS, which is a fork of the Mozilla TTS project with a web server frontend. We again enable the server on boot with systemd. From here, you can send the server strings to have it spoken as audio data.
Finally, we configure Mycroft AI to use these two servers rather than its default STT and TTS services.
Mycroft still requires a remote backend to enable various skills. While the backend is open source (github.com/Myc..., it is a pain to set up. We will save making Mycroft fully offline for another time.
Product Links:
Raspberry Pi 4B: www.digikey.co...
Related Videos:
Jayy’s companion bot with Mycroft AI:
/ 1495921164497076224
How to create a custom skill for Mycroft AI: • How to Create a Custom...
How to create a custom wake word for Mycroft AI: • How to Create a Custom...
Related Project Links:
How to create custom STT and TTS servers for Mycroft AI: www.digikey.co...
Related Articles:
How to create a custom skill for Mycroft AI: www.digikey.co...
How to create a custom wake word for Mycroft AI: www.digikey.co...
Learn more:
Maker.io - www.digikey.co...
Digi-Key’s Blog - TheCircuit www.digikey.co...
Connect with Digi-Key on Facebook / digikey.electronics
And follow us on Twitter / digikey

Опубликовано:

23 сен 2024

Ссылка:

Скачать:

Готовим ссылку...

Добавить в:

Мой плейлист

Посмотреть позже

Комментарии : 14

@citizen240 2 года назад

Excellent presentation on setting up a local TTS/STT. BUT, because of the pace, the precision of language (no “this dohicky”, etc.), all information clearly displayed on screen, and especially showing the resolution of errors, it’s also a very condensed master class on just plain getting stuff done in a Linux environment. Decades ago I sweated through the pain of setting up CAD/CAM applications on IBM RISC workstations running AIX, getting acclimated to arcane, lengthy, case sensitive commands. I have bookmarked this video as a great resource for navigating in the current Linux world. Thanks a lot! Much appreciated!

@TradieTrev 2 года назад

This is really cool! I really love this type of content Digi-Key!

@TradieTrev 2 года назад

Had mixed experiences with installing software, It's always a mind field having to deal with different software versions and needing hardware support. Thanks for sharing your efforts!

@keithlambell1970 2 года назад

Thank you. An excellent walk-though.

@danielash6929 2 года назад

The voice commands are a must-have but there languages are gearing up.

@akissot1402 Год назад

Why you didn't use "Mimic 3" from mycroft ? the site reads "In human terms that means it sounds great and can run completely offline on hardware you control." is that offline robotic quality too ?

@akissot1402 Год назад

i want these glasses god d.. it! i can't find anything like it just by googling, i would wear them even in the club

@ftab 2 года назад

This is really cool. If the STT/TTS server has to run on a high end processor, is there any reason the Mycroft client couldn't be on an ESP32 or something like that?

@danielash6929 2 года назад

Yes you are right the mosfets could share datas in fer red. Cross talking catch each others program.

@ShawnHymel 2 года назад

I had the same thought. In theory, yes. For example, you could send a string in a packet to the TTS server and have it respond with audio. You could then play the audio out over, say, an I2S speaker. :)