"ReSpeaker Lite" - new Seeed Studio Voice Assistant Development Kit hardware combine ESP32 with XMOS XU316 DSP chip for advanced audio processing as a ESPHome-based Home Assistant Assist Satellite voice devkit

formatBCE · February 23, 2026, 2:43pm

You should understand difference between OWW and MWW.
OWW is operated on server, so your satellites will ALWAYS stream audio to server, if there’s slightest noise in the room. If you have several of them, you also will notice raised CPU usage on server.
MWW is on-device engine, so before wake word is recognized, it’s completely device-local, no streaming audio.

So far I haven’t seen useable configs with Openwakeword for VPE/Respeaker. There’s too much to change for it to work, and things like “stop” word will be lost.

Try using different word, or train your own model. For me “Ok Nabu” works pretty good.

Skyman1 · February 25, 2026, 8:53am

It seems like it’s not the best time to buy new hardware… prices have increased by 40% at the moment.

Hedda · February 25, 2026, 11:37am

Tip is to try to find a used mini-PC on local private market places if possible.

Skyman1 · February 25, 2026, 3:26pm

I don’t know, I’m thinking about it but maybe it’s better to wait until this summer where I’m convinced prices will drop…
second hand goods have also increased by 40%, if not 50%. Not that I have money problems but it bothers me to give them away.

Hedda · February 25, 2026, 8:55pm

This is getting really off-topic and hate to be the carrier of bad news + do not want to discuss this further here, but FYI, from all what I read about it (which is a lot from many independent sources because affect me at work too as well as my hobbies) is that the RAM and Storage prices are not likley to go down until the year 2028 if the AI-bubble does not pop or deflate before then, so prices on the used computer market are not likley to drop before then either as it is directly affected. Hence I am sure it will get worse before it gets better and the prices will get even higher this year before possibly leveling out next year (again depending on if the AI-bubble does not pop or deflate before then). For quick reference see:

formatBCE · February 25, 2026, 9:11pm

*if it gets better. Manufacturers are easy on going price-up, but not that much on pricing down…

Skyman1 · March 5, 2026, 9:30pm

*wake word with Android was released…
do you think an Android device can replace all these devices that work as assistants?

It has an audio output and probably a microphone and software that can overcome all the problems of false activations… even a tablet connected to a Bluetooth speaker.

formatBCE · March 5, 2026, 9:43pm

If you have it on always-powered device.
And still - I have ported Openwakeword to Android. It’s fine, but not outstandingly great.
Also it’s much more expensive devices than Respeaker.

Skyman1 · March 6, 2026, 6:26am

Honestly, I don’t find the Respeaker lite “outstandingly great”, maybe the 3800 is better, but it has a similar price to that of a low-end tablet or smartphone (we often find ourselves with old smartphones or tablets unused at home).L

formatBCE · March 10, 2026, 1:30am

Oh, by “outstanding” I mean “better than we usually see now”. It’s not.

But maybe it will work for you.

Skyman1 · March 10, 2026, 9:53am

I think a smartphone microphone works much better… or maybe just a very expensive phone works better.

adriansinho · March 25, 2026, 11:47am

I am experiencing an issue with my ReSpeaker Lite device where it seems to continuously trigger listening even without saying any wake word.

I am using the “Okay Nabu” wake word.

From the logs, it looks like the wake word is being detected randomly, and the voice assistant pipeline starts (STT, microphone streaming, etc.), but no actual speech is provided. This results in repeated stt-no-text-recognized errors.

Additionally, it behaves as if the physical button is being pressed intermittently, even though no one is touching the device. This triggers actions and sounds unexpectedly.

There is also a constant “ping” sound (notification sound) happening repeatedly, which makes the device quite noisy and unusable.

My setup:

Firmware version: ver=0110
ESPHome version: 2026.3.1
YAML configuration: using the latest version from the repository (2026.3.0)

Has anyone experienced similar behavior? Could this be related to wake word sensitivity, VAD configuration, or possibly noise/interference triggering false positives?

Any guidance or known fixes would be appreciated.

formatBCE · March 25, 2026, 8:07pm

Did you solder USR button as it’s said in instructions? Without that it sporadically calls clicks on it, because that pin is floating by default.

adriansinho · March 26, 2026, 9:22am

I didn’t see anything about that in the documentation.

However, based on this hint, I changed the button configuration to the following:

binary_sensor:
  - platform: gpio
    id: user_button
    pin:
      number: GPIO3 # D2
      inverted: true
      mode:
        input: true
        pullup: true
    name: "User button"

It is now working properly.

Thank you very much for your help.

formatBCE · March 26, 2026, 12:40pm

There’s section with literally this text in main README:
ATTENTION! This step is mandatory, as without it the buttons on satellite won’t work as intended.

Whatever works for you, but for working buttons they should be soldered to corresponding pads. Without that you can’t use them to pause music, stop ringing timer, or call the assistant without wake word. Same with programmatic mute control.

jsbrich · May 4, 2026, 5:46pm

I just got one of these and I saw that in the documentation, but wasn’t sure exactly what it meant. Do you just solder a jumper wire between the two holes ie Mute and D3 and D2 and User?

formatBCE · May 6, 2026, 5:39am

Yes, for everything to work correctly. Jump-wire them by pairs.

photo64 · May 12, 2026, 1:12pm

I decided to update my respeaker-lite satellites last night to the current esphome version (2026.4.5) and I saw there was a recent update for the common.yaml file (2026.3.0) so I also updated this file on my system. The 1st one I did work without issue however the 2nd one I attempted to update seems to be stuck in a loop (I think). It seemed to upgrade ok but it won't start or respond. I tried to do a manual file upload but the same thing is happening. I am currently running HA - 2026.4.4.

I didn't see any known issues regarding a problem but was wondering if anyone has run into this upgrading to esphome 2026.4.5?

Any ideas on how I can recapture this device? I don't think it's bricked, looking at the logs I think it's stuck in a loop. I see an assert failed Line then it repeats this:


Backtrace: 0x40380fad:0x3fcead30 0x40380f75:0x3fcead50 0x403854a9:0x3fcead70 0x420b3cb5:0x3fceae90 0x420b3dd9:0x3fceaeb0 0x4210b53e:0x3fceaee0 0x4210b599:0x3fceaf20 0x4200d93d:0x3fceaf40 0x4200d989:0x3fceaf60 0x4200d9de:0x3fceb190 0x4209c68e:0x3fceb1e0 0x4209c8ad:0x3fceb200 0x4209c361:0x3fceb220 0x4209c286:0x3fceb240 0x42097cb9:0x3fceb260 0x42097ccc:0x3fceb2a0 0x42099c2b:0x3fceb2c0 0x42099c5e:0x3fceb2e0 0x4037a5ca:0x3fceb310 0x403ccc54:0x3fceb340 0x403cd0f1:0x3fceb380 0x403c8981:0x3fceb4b0 0x40045c01:0x3fceb570 0x40043ab6:0x3fceb6f0 0x40034c45:0x3fceb710
[22:20:17]
[22:20:17]
[22:20:17]
[22:20:17]
[22:20:17]ELF file SHA256: 0a0d816d2
[22:20:17]
[22:20:17]Rebooting...
[22:20:17]ESP-ROM:esp32s3-20210327
[22:20:17]Build:Mar 27 2021
[22:20:17]rst:0xc (RTC_SW_CPU_RST),boot:0x2a (SPI_FAST_FLASH_BOOT)
[22:20:17]Saved PC:0x40381155
[22:20:17]SPIWP:0xee
[22:20:17]mode:DIO, clock div:1
[22:20:17]load:0x3fce2820,len:0x152c
[22:20:18]load:0x403c8700,len:0xce4
[22:20:18]load:0x403cb700,len:0x2f48
[22:20:18]entry 0x403c8918
[22:20:18]I (24) boot: ESP-IDF 5.5.4 2nd stage bootloader
[22:20:18]I (24) boot: compile time May 11 2026 21:49:18
[22:20:18]I (24) boot: Multicore bootloader
[22:20:18]I (24) boot: chip revision: v0.2
[22:20:18]I (25) boot: efuse block revision: v1.3
[22:20:18]I (25) boot.esp32s3: Boot SPI Speed : 80MHz
[22:20:18]I (25) boot.esp32s3: SPI Mode       : DIO
[22:20:18]I (25) boot.esp32s3: SPI Flash Size : 8MB
[22:20:18]I (25) boot: Enabling RNG early entropy source...
[22:20:18]I (26) boot: Partition Table:
[22:20:18]I (26) boot: ## Label            Usage          Type ST Offset   Length
[22:20:18]I (26) boot:  0 otadata          OTA data         01 00 00009000 00002000
[22:20:18]I (26) boot:  1 phy_init         RF data          01 01 0000b000 00001000
[22:20:18]I (27) boot:  2 app0             OTA app          00 10 00010000 003c0000
[22:20:18]I (27) boot:  3 app1             OTA app          00 11 003d0000 003c0000
[22:20:18]I (28) boot:  4 nvs              WiFi data        01 02 00790000 00070000
[22:20:18]I (28) boot: End of partition table
[22:20:18]I (28) esp_image: segment 0: paddr=00010020 vaddr=3c130020 size=15ff0ch (1441548) map
[22:20:18]I (283) esp_image: segment 1: paddr=0016ff34 vaddr=3fc9b800 size=000e4h (   228) load
[22:20:18]I (283) esp_image: segment 2: paddr=00170020 vaddr=42000020 size=12d7ech (1234924) map
[22:20:18]I (501) esp_image: segment 3: paddr=0029d814 vaddr=3fc9b8e4 size=04de4h ( 19940) load
[22:20:18]I (506) esp_image: segment 4: paddr=002a2600 vaddr=40378000 size=137b8h ( 79800) load
[22:20:18]I (523) esp_image: segment 5: paddr=002b5dc0 vaddr=50000000 size=00020h (    32) load
[22:20:18]I (531) boot: Loaded app from partition at offset 0x10000
[22:20:18]I (532) boot: Disabling RNG early entropy source...
[22:20:18]

formatBCE · May 12, 2026, 1:34pm

The stack trace is useless, it needs to be bound to exact addresses in your device memory to be read.

Try cleaning .esphome directory with caches, rebooting ESPHome and recompiling. Usually it solves the problem. Also install it via USB.

photo64 · May 12, 2026, 1:38pm

Could you explain what you mean by "cleaning .esphome directory with caches"? I have been trying to install via USB and doing a "Clean Build" prior to compiling