"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

As usual super glue seems like the solution to my main issue. I just don’t like it falling off and yanking on the JST wire, especially after that post above. Super Glue and duct tape can fix almost anything although duct tape tends to lose some style points unless hidden.

1 Like

Omg after multiple flashs, removal from esphome integration and the lot it finally works. woo hoo!!!

The performance is still crap though. My fault for expecting miracles. lol

Sorry dont mean to be nasty or anything. Just an observation I guess. I havent used it with local AI which was taking too long in my initial tests. Might try that.

Also I had to enable the respeakerv3 led to make it work after the flash. The seeed code it was working on its own.
image

Hi!

Does anyhave have the JTAG pinout of the ReSpeaker Lite available?
(We really need the schematic. Don’t want to find all the details by manually measuring.)

I plan to use my XMOS XTAG4 adapter to experiment with custom builds (I2S/UA variants) of the XMOS firmware from here with 48kHz enabled instead of the default 16kHz.

One could try that without a XTAG4 adapter, but I feel more comfortable when I am able to reflash in case of failing firmware images.

The design files of the XTAG-4 adapter can be found here. Fun fact: it also includes a XU-316 chip and could be reflashed.

1 Like

Great news!
Maybe you could open up another thread for further discussions?
I have a couple of questions. Especially about the used XMOS processor and its firmware, because you are using 4 mics instead of only 2. :slight_smile:

New thread created :slight_smile:

You use some other YAML I guess? There’s no switch exposed in mine. IDK what that is, so probably it’s still doesn’t work as expected…

I would LOVE to see 48kHz! :slight_smile:
Have no idea on JTAG pinout, but I think it’s worth trying to ask on Seeed Discord.

That’s the listen light LED. That’s why it’s not enabled by default. Even though my listen LED doesn’t work if I expose that and say the wake word “ok nabu” it turns on and changes colors as it progresses along with the voice pipeline. When the command ends it turns back on. Seems to emulate the on board LED but since mine doesn’t work then I can’t verify but pretty positive that’s all it does.

Are we all still on the same page? I didn’t expose the LED in config, dod you do it to test?

Yeah exactly what I was thinking.

Question: Does your perform poorly as well?

Doesnt seem like this is better than another one I built with esp32-s3-devkitc-1(N16R8).

It’s way better for me. I have 3 S3-based satellites with INMP mic.

Interesting. Could it be the host/VM spec?

I am on a 13th Gen 32GB no graphics card Win 11 Pro system.
Home Assistant on VM player with basic settings.
2 CPU/2 GB

I don’t think it matters really. Although STT can give difference. But it depends on where you have troubles: wake word, speech-to-text, or noise cancelling.

I received my ReSpeaker Lite Kit yesterday.
All is working, except the RGB LED isn’t, or barley.
When using the Arduino LED example, it shows the initial colour set (all colors work fine) but then only a power cycle helps to set a new color.

GND & VCC is fine, but I measure like 300 kOhms between GPIO1 and DI of the LED, seems a little high. There is activity on GPIO1 when the LED should be ON, will hook up my Oscilliscope later. Could anyone with working LEDS quickly verify what resistance is between GPIO1 and DI?

image

I measured mine and have pretty same value. ~400kOhm. My LED is working.

1 Like

However, i have another board, to which i soldered S3 myself. Has +5V across and working connection DIN-GPIO1 with same resistance. But LED doesn’t work.

I have no idea why but yes, using your yaml it shows up as disabled by default, If I enable it it goes through the voice pipeline colors in the UI but the LED still doesn’t work. I have zero idea why it’s but it does. I just cleaned the build files and installed after deleting the integration to be sure.

I like the new little on boarding process they have. Not sure when that was added. After saying the wake word twice, then the last pic comes up. After that it prompts you for the voice pipeline to use so it does everything you need, especially if you have multiple voice pipelines like I do. Since they removed the binary sensor it’s still there but hidden by default and “Assist Satellite” takes it’s place.

Personally, I think it works better then the S3 box when not using the touchscreen button on the S3 to hold and let go to stop listening. Not sure if that could be done with the USR button on the respeaker lite board or not but I would say it’s a close second to my dedicated speakerphone plugged into my HA server using Assist Microphone. To my understanding that add on is supposed to at least utilize some of the DSP tech in whatever USB device it’s plugged into. I would say it’s equal to a wyoming satellite with room to improve over time. Maybe not this particular model, but future versions, like nabu’s and others.



Yeah, it’s pretty cool device. Onboarding process is nice, but for some reason it removed my TTS from my pipeline after setup, so I had to set it again (after several minutes of confusion why my satellite does work but doesn’t speak).

I got 2x bare boards too, and tried to solder my own XIAO S3 to one of them. S3 works, Respeaker works via USB with Mac - but doesn’t work with board. Triple checked LED connection to GPIO1 and 5v/gnd - but LED doesn’t work too…

I’m trying to get to Seeed support, but they stopped responding… Shame. I’d gladly have 2 more good satellites for my home.

Hi Guys,

First of all forgive me for my ignorance I am new to all this stuff!

So from what I understand.

  1. I have to build ESPHome for the S3, this then uses a YAML file to work out how to talk to the audio side of the board.

  2. Currently it is only 16K in and 16K out, but it looks like there is the possibility of 48K out in the future.

Is the best YAML to use from @formatBCE : Create esphome-respeaker-lite.yaml · formatBCE/Random-snippets@5eef002 · GitHub ?

I’m guessing there is some kind of server needed somewhere as well (Home Assistant)? I also need to install this on a PI?

Am I on the right track here or am I hopelessly confused :slight_smile:

Thanks

Andy