Running Jasper (interactive speech) on Raspberry Pi Zero W with Pimoroni SpeakerPhat
I was really interested to get as compact an implementation as I could of an interactive speech app such as Jasper or Alexa, using Rpi Zero and a built-in mic and speaker. I had a Pimoroni SpeakerPHAT laying around from earlier playings-around, and that seemed perfect for the task. Here's what I ended up with, and while I'm just using my USB battery here, to make things more compact I could equally use the flat battery listed in my Jupyter/Rpi page.
The parts here are available from Adafruit.com:
Starting with the punchline first, here's a 1-2min video in which I'm playing with / demonstrating how (and how well) Jasper works on there. Being used to hearing Siri and Alexa and Google's agent has me sortof spoiled in terms of expected level of quality - this thing is open source, is a community effort, can be installed entirely locally and is pretty configurable/personalizable; and for all that it's pretty neat. But it's no Siri or Alexa or Google agent by any means. No judgement meant toward the developers there, I still think it's really neat.
I was really interested to get as compact an implementation as I could of an interactive speech app such as Jasper or Alexa, using Rpi Zero and a built-in mic and speaker. I had a Pimoroni SpeakerPHAT laying around from earlier playings-around, and that seemed perfect for the task. Here's what I ended up with, and while I'm just using my USB battery here, to make things more compact I could equally use the flat battery listed in my Jupyter/Rpi page.The parts here are available from Adafruit.com:
- Raspberry Pi Zero W ($10),
- SD/MicroSD Memory Card (8 GB) ($10),
- Mini USB Microphone ($6),
- SpeakerPHAT by Pimoroni ($13),
- USB battery such as this one ($15).
Starting with the punchline first, here's a 1-2min video in which I'm playing with / demonstrating how (and how well) Jasper works on there. Being used to hearing Siri and Alexa and Google's agent has me sortof spoiled in terms of expected level of quality - this thing is open source, is a community effort, can be installed entirely locally and is pretty configurable/personalizable; and for all that it's pretty neat. But it's no Siri or Alexa or Google agent by any means. No judgement meant toward the developers there, I still think it's really neat.