To add on to what the other replier said, these two projects you wish to combine are not written with that intent. They are the root firmware, effectively.

Modular systems like ESPHome are written with that modularity in mind, and have underlying code APIs (not like a REST endpoint) that those modules hook into and utilize.

Even if WLED and Espectre were written to be modular firmware, they would not be compatible with each other, nor ESPHome, unless expressly written for such.

As far as the “calling Github” part: Yes, sort of. It’s meant for making it easier to pull modular packages others have written for ESPHome to share with others, allowing easy inclusion of code that doesn’t require redownloading files yourself every time they update it for changes in ESPHome or bug fixes.

What do you mean by “local”?

If you mean "on my network, never going to the cloud, install whisper and piper apps (in default repos at that) in Home Assistant OS: you have a fully local voice stack that multiple assistants can use. If you might have multiple speakers, why duplicate resources for processing speech?

HA basically compartmentalizes voice into into its parts: speech to text, intent processing, and text to speech. It’s fully local if you want, or cloud, or both, entirely up to you.

If you mean “this software processes voice itself from the attached microphone, interprets the command, sends the command to HA, and then creates a TTS response or chime”, there are software for that but they certainly aren’t running on an inexpensive ESP32-S3 chip, and are generally gonna be pretty limited as well, and I don’t see active projects using it anymore. Everything that was just…tended to work poorly.

I believe I’ve seen some self-contained software packages that can connect to HA’s Assist over an API, feeding it the transcribed text directly, but again, a central source of processing speech on more-powerful hardware that is only ever going to be used like, one user/speaker at a time makes way more sense than duplicating efforts on worse hardware.

I don’t feel like I fully saw this explained yet:

Home Assistant itself is what provides the voice “stack” in the form of “Assistant” definitions. You can have multiple, which require defining a speech-to-text provider, a text-to-speech provider, and a langauge processing provider (agent).

By default for that last one it’s Home Assistant, but LLMs can be selected too. The STT and TTS providers can be Home Assistant Cloud, which uses Azure Speech for both services. You can also install (Faster-)Whisper and Piper add-ons/apps for STT and TTS, respectively, in HA OS. (Or run them on a separate system, like in my case I use the GPU in my server since HA runs on its own micro PC).

This allows defining multiple voice “stacks” at once, so you could, for instance, have one wake word on a given device be for one pipeline for home commands, and another that just talks to an LLM.

Crucially, these do not replace the voice stacks for a Google Home/Nest or Echo device (without the hardware replacements mentioned). Those still go to their respective cloud servers for processing, not locally.

The Voice PE device is largely referred to as a “voice satellite”, which is apt – it is an extension of sorts to your existing stack. It is a device you can put somewhere convenient, like a bedside or coffee table, that is like a voice connection, sending your voice to Home Assistant for Home Assistant to process using the selected means for the pipeline active, and respond in TTS the same. It’s a microphone and speaker with wifi, effectively, and that’s all, as others have said.

I’m not sure I follow what you mean by “premium” though. The Voice PE device, and similar, are ~$50-60 smart speakers just like Nest/Echo (which again also rely on external processing). If you mean having Home Assistant Cloud as a subscription service, then yes sort-of – that would give you a TTS and STT provider.

That’s kind of irrelevant for a third party resolving where network traffic is coming from. They aren’t connecting to OP, OP is connecting to them. Assuming their firewall even supports reverse DNS lookups, it’d actually resolve to something like <Ipaddress>.res.ispdomain.com or something weird. And since it’s dynamic there’d be no consistent domain to give.

Can you share photos of the underside of the remote and/or light where it lists either “FCC ID” or “Contains FCC ID”? I can’t find a listing on it anywhere to figure out what it does, and a LOT of those labels for radio frequency in Amazon images are straight up lies. There’s so much noise on those drop-shipped items.

I do all three every time I do something new, it’s basically mandatory for me. Every time I did it correctly did numerics tested over and over via voice itself. Timers always behave and require that as well.

This was just the first time, after anxiety had faded, it had failed to transcribe in a useful fashion.

Makes me want to look into more-predictable STT systems.

I haven’t had disk issues, am running a 980 Pro SSD currently, but I’ve definitely noticed other weirdness that sure feels more like that? Half-Life crashing repeatedly in map loads sometimes succeeding fine and other times not. Firefox broke wholesale until I reinstalled it and even then had to do a refresh to fully solve it. I haven’t seen anything else weird thankfully but this definitely has me concerned and glad I’m backing up with a very long rolling period just in case. Gonna uninstall this update for sure.

My assumption was it had something to do with the drivers for the controller, and the update flooding it faster than it could take data and not caching everything it couldn’t cram into controller DRAM, causing parts to just get dropped wholesale.

If you want to argue in favor of your slop machine, you’re going to have to stop making false equivalences, or at least understand how its false. You can’t make ground on things that are just tangential.

A computer in 1980 was still a computer, not a chess machine. It did general purpose processing where it followed whatever you guided it to. Neural models don’t do that though; they’re each highly specialized and take a long time to train. And the issue isn’t with neural models in general.

The issue is neural models that are being purported to do things they functionally cannot, because it’s not how models work. Computing is complex, code is complex, adding new functionality that operates off of fixed inputs alone is hard. And now we’re supposed to buy that something that creates word relationship vector maps is supposed to create new?

For code generation, it’s the equivalent of copying and pasting from Stack Overflow with a find/replace, or just copying multiple projects together. It isn’t something new, it’s kitbashing at best, and that’s assuming it all works flawlessly.

With art, it’s taking away creation from people and jobs. I like that you ignored literally every point raised except for the one you could dance around with a tangent. But all these CEOs are like “no one likes creating art or music”. And no, THEY just don’t want to spend time creating themselves nor pay someone who does enjoy it. I love playing with 3D modeling and learning how to make the changes I want consistently, I like learning more about painting when texturing models and taking time to create intentional masks. I like taking time when I’m baking things to learn and create, otherwise I could just go buy a box mix of Duncan Hines and go for something that’s fine but not where I can make things when I take time to learn.

And I love learning guitar. I love feeling that slow growth of skill as I find I can play cleaner the more I do. And when I can close my eyes and strum a song, there’s a tremendous feeling from making this beautiful instrument sing like that.