Home Assistant Voice Assistant with local Ai
Home Assistant Voice Assistant with local Ai
It has been a rough start to getting a good quality echo replacement but Home Assistant is amazing. Some parts got silly enough to need a video. Using Ollama as a local Ai server has been working much better vs the native devices capabilities, but wow this little ESP32 is not doing great. I also think I made some other mistakes along the way that I am going to need to fix.
M2000 GPU https://geni.us/Quadro_M2000
ESP32 Box (for tinkering imo) https://geni.us/ESP32-S3-BOX-3
Zimablade https://geni.us/zimablade-3760
Build a Local Ai Server for $350 https://youtu.be/iflTQFn0jx4
Be sure to 👍✅Subscribe✅👍 for more content like this!
Join this channel https://www.youtube.com/@DigitalSpaceport/join
Digital Spaceport Website https://digitalspaceport.com
Patreon https://www.patreon.com/digitalspaceport
Buy Me a Coffee https://buymeacoffee.com/digitalspaceport
Please share this video to help spread the word and drop a comment below with your thoughts or questions. Thanks for watching!
Chapters
0:00 Voice Assistant in Home Assistant
2:06 ESP32-S3-Box-3 Review
3:30 Philips Hue in Home Assistant
4:08 Is the S3 Box good for anything?
5:35 RPi4 as Home Assistant Voice Ai Server
7:05 Planning Smart Home Layout
8:06 Old smart devices are okay?
*****
As an Amazon Associate I earn from qualifying purchases.
When you click on links to various merchants on this site and make a purchase, this can result in this site earning a commission. Affiliate programs and affiliations include, but are not limited to, the eBay Partner Network.
*****
Im kike 3 minutes in and this is already your funniest video yet
see networkchuck full homeassistant Ollama ai home automation
he did it good
sarakit for the pi compute modules looks like it was built with this specific purpose. 3 mics and some kind of extra processor. suspiciously round.
i almost ordered the esp box s3 when i saw the willow voice assistant being developed for the hardware. i wonder if the willow devs squeezed more performance out of the Box.
Thanks for the honesty. I also tried to copy network chucks HA voice demo, and found the results to be disappointing in real life scenarios.
Amazon is coming out with Amazon Plus to upgrade their rudimentary speech driven AI… for an extra 10 month cost. After having Alexas for years, after that initial Wiki walk like fever wears off asking them about things they are good about like chemistry or mythology (and not good at, like computer hardware), mostly you end up just using them to turn the lights on and off, check the weather, set reminders and timers, check the date or time, or do calculator math. So I don’t think I’ll be jumping on Plus, unless it is absolutely fantastic or could tie in with my home rack computer systems.
Hey mate, just found your channel and subscribed!
I wanna see more homeassistant content!!!
Heck ya! I’ve really wanted to get a voice interface working for HASS. I do wish someone would start selling a nicer hardware for this.
Also a dashboard for HASS is needed; I’m using 11" android tablets but it would be nice to see dedicated hardware for this. Thanks for the vid!
after going full smart home, the dumbing down of Alexa ended me and I’ve returned to a biologically intelligent activated lighting and sound system (aka a manual switch).
Looking forward to your mid-tier and other local llm server projects, for those of us on a budget – can afford more than 600 for GPU, but not 3500 for 4090…
I’m interested in how the smaller models perform compared to the online models (claude and o1 etc)
great video.. but for some reason it won’t let me link to the HA community custom firmware.. it improves the s3-box by a *huge* amount.. and yes.. even makes the speaker useful again!.. just google "ESP32-S3-Box3-Custom Firmware"
More HA 😂 I’m just now getting into this so more videos please 😅😊
those that bought those google or amazon or apple virtual assistive devices in my view actually have lost not only money but time, privacy and ability to impact the world of ai based assistants and home automation in general. now to road back to privacy focus, locally ran ai and home automation is going to be slow and potentially impossible at least for most.
I uh tried Ollama Integration for Home Assistant, Nemotron was able to make sense of my commands somewhat but it’s a looong way to go. You need local Whisper pipeline, TTS, STT and so on. Can’t wait to see what kind of open source hardware (esp. good long distance mics) we about to get soon!
NVIDIA Jetson Orin ??? Runs ollama good
I agree about Alexa going downhill
Have you tried the Home Assistant Voice Preview?
The first and only Alexa device I owned is literally inside a wall because I didn’t even bother to fish it out before repairing the Echo Dot shaped hole in the drywall. I don’t have fits of rage…. Except that one time.
It would be great if someone figured out how to use Amazon Echo and Google speakers as microphones and speakers for HA.
I too agree with Alexa going downhill. It could be the cloud is so crowded that there are mishits all the time. I also agree that most people will find that HA voice command (local control) will only find it initially about 30% as the video explains and only after understanding how it works and playing with it. HA 2024 is the year of the voice. It’s still very new but I can see it improving in leaps and bound. Before voice command, it only needed trivial processing power. Now, in order to keep it local, and I got rid of any device that could not operate locally too, you suddenly need access to this huge processing power and memory compared to what worked before. I had switched to a Optiplex Micro I7-9700S PC which is pretty quick but even with this, the voice model could not be larger than "small" without getting appreciable delays. The "medium" model had a 4-5 sec delay, too slow. So, I have decided to go big guns and get a i7-147000S with massive processing power. I won’t know the improvements until I get it in but there was a huge difference rom the I5 to I7 model. I am already assuming that I will be able to use at least the "distilled medium model" and possibly the a "large" model which should increase accuracy very significantly without significant delays … hopefully! The Optiplex Micro 7020 i7-14700 ain’t cheap!!! PS: you can use a USB speakerphone with HA and that overcomes the almost inaudible voice but be careful, it must be completely plug and play like the Kay/Uda and that was not working until last year. It still takes HA awhile to recognize it. Installed using Assist Microphone in HA.