My wife is in China so I have access to buying there for the next few weeks. Prices are so much lower that I’m scrambling to order things I think I want before the window for her to carry them home to the US for free closes.
Adding voice to HA so I can reduce dependence on google is on my short list along with DIY occupancy sensors. Parts for the occupancy sensors are on the way and now I’m looking at voice.
The price of the S3 box devices is higher than I want to spend to get started, but the Atom echos are low enough that I’m willing to give them a try.
My hopes for an initial try is to enable basic announcements such as “garage door is open” and simple commands such as “close garage door”. Assuming I can get these working I would position the echos at strategic locations around the house.
For reference, I’m running HAOS on a Beelink S13 with 16GB of ram. I have some wifi and zigbee devices, a half dozen Cinnado D1 cams flashed with Thingino, a Reolink doorbell cam, and I recently added a Reolink NVR with four cameras attached so far. At this point, the Beelink is far from overloaded.
This leads to a few questions.
First, I think the Echo Atom route is a good start. Is this reasonable given my initial goals or am I going to waste time and money?
Assuming this a decent starter path, does it make sense to buy a handfull of them now since I can get them at so much better price than in the US?
This is a very hands on route - fine if you like tinkering. Also, you’ll get what you pay for - for practical use I would expect you to have to add another speaker of some sort, so it won’t be as cheap as you think.
There’s lots of posts about this - for example:
Do a bit of reading first.
Edit: I have several S3 Boxes and the sound quality isn’t great there either. If you’re coming from Google you (or your family) may be disappointed.
I’ve been a tinkerer for quite a while with both hardware and software. I built a Christmas light display for my front yard which uses strings and strips of RGB lights. It is controlled by four boxes which contain a power supply and a nodeMCU running WLED. I use xlights to synchronize them to music. I’ve also pulled lots of wiring in my house and installed and configured the alarm system.
Agreed. I need to do lots of reading.
My expectations for these are low. I don’t plan to use them for music, just fairly basic voice output. Background music will stay on the google devices or one of the surround systems when I want more volume.
I think buying a one or two to learn with is reasonable… but expecting them to be replacements for any of the Google smart speakers is unrealistic.
Remember these things are tiny, with a tiny speaker and low power… so they can’t get that loud without distorting the audio. You’re not going to hear it clearly from another room like you can with more powerful smart speakers.
I use the two I have as incidental voice input devices for rooms where:
There isn’t much background noise.
I occasionally want to issue a voice command, but not so often that it’s worth putting a full power smart speaker there.
Before you go for a shopping spree just keep in mind that as for line voltage devices the standard in China is 220-240V AC power with a frequency of 50 Hz while the standard in the US is 120V and 60Hz AC.
I have one I used for my early voice experiments. It’s in a drawer. If you live in Texas come by you can have it. Not worth my time to box it up and send it to anyone so no I’m not doing that. I have 6 VPEs now.
Without external amplification and a speaker you will be VERY disappointed. They’re good for testing or small close low noise environments like a single person’s desk. IMHO.
I know it’s tempting but by the time you bring home your haul and have spent 100 bucks on stuff you’ll never use. Did you really save anything?
I have 1st gen atom echo. Bought usb speaker pair and separated them to install an echo in each(basically upgrading echo to 4” speaker). $15USD for 2 speakers and they sound Amazing.
I placed echo in a manner that allowed mic to function but they are mostly for announcements.
The new echo is esp-s3 I belive making it sendspin capable. I would definitely do this again presuming under $20USD for wireless speaker(voice control not needed)
Interesting option. I did some quick searching and it looks like I can buy a pair of USB speakers and two echos for around $32. Did you put the echo inside the speaker whole or just the guts?
Desoldered echos speaker and solder usb speaker in its place (make sure you get polarity correct or sound will be weird)
The speakers I used had slot that allowed esp to fit so I cut a spot for it leaving usb power port exposed. If fit tightly so when reassembled oiliness solid fit. Only issue is led not very visible but wake word works and voice works.
Sound was good. I just wanted cheap speaker to spread around house for announcements and initial transition from Amazon. I had a bunch of echos that sounded terrible and tried this and worked great. Bathroom, laundry room, places were announcements and music is useful but voice is not necessary. Also can serve as supplemental speaker in room that already has main voice device providing surround. I didn’t expect much but easily filled 12x12 room with sound equal 3rd gen Amazon echo(I think, didn’t compare but it was good).
I used these
Will add picture
Mounted in back of base
Thanks. I just remembered that I have an old pair of PC desktop speakers I can try. I have about a two week window to get a prototype working before the window closes on being able to carry home parts from China.
If I can get them working with an echo I’ll enlist my daughter to design and 3D print cases.
Assuming the first batch goes well, the speakers you posted look good for building additional devices.
Standard atom echo is not sendspin capable I believe. New (aqua/blue) echo has esp32-s3 with psram so it should work with sendspin. Not a make or break but if you plan to pair with another device you’ll likely want sendspin
It’s interesting how many new things to think of can come up. I started out thinking I just wanted a device to hear commands and output simple messages. After reading about sendspin, whole house audio without the exorbitant prices some of the manufacturers want seems like it could be possible sometime if the future.
Your post also resulted in me wondering what the heck is psram and how does one know if the board has it?After a bit more searching and reading, I know a little more. I already had some esp32-s3 N16R8 boards in my taobao cart, if I read right, they have 8MB of psram.
The additional digging also makes me wonder if I even need to buy an atom if I’m just going to gut it? It seems like I just need to choose an appropriate board (esp32-s3 N16R8 ?), amplifier (MAX98357A ?), microphone (INMP441 ?), and speaker and I’ll have the necessary hardware. Wiring them up is not a concern, the challenge will be on the software side…
In the interest of taking small steps, I need to stay with my original goal of simple commands and voice output, but doing a bit of future proofing with good hardware choices makes sense.
Even not being that expensive: if you plan to make modifications, don’t buy.
You can get an ESPS3 board for a lot less.
I have one because I was curious how voice would work but after some time I got a VPE and even being more expensive: it’s worth it’s money.
I’m also using a repurposed tablet with voice and that’s even better compared to the echo (definitely the speaker but also the mic)
The ability to detect the wake word is way more sensitive on the VPE and from the 3 satellites I use, the echo is the worst.
When there is background sound (TV, music) the difference is even bigger.
My experience was almost identical, Nick and yes 100% agree. If you’re just going to gut it. Don’t. In fact I’m very big on don’t buy until you identify the exact need else it’s easy to throw away cash tbh.
Got it. When I posted it seemed like a good thing to start with, but based on the input I’ve received and additional reading, echo is off the list.
The VPE clearly is a good choice, but I’m thinking of deploying in at least 10 locations for announcements and basic commands and using VPEs would be too much money. Starting out with a small number of ESPs is not risky because I can repurpose them if I decide on something else later. If they meet the initial goal, it will have been worth the small amount of $.
It sounds like microphone choice and processing is important. I do have a few old tablets I may try to use later.