Has anyone out there hooked HomeAssistant up to a local LLM? I'm very tempted:

• Alexa integrations fail often. HomeAssistant integrations tend to be rock solid.
• Alexa is rule/pattern matching based. LLMs can understand natural language fairly well. The "magical incantations" required by Alexa are awkward.

Other than the software, the device side seems challenging. There are $50 fully-baked POP devices. I'm less sure on the DIY front.

Also, I desperately want my house to speak to me in the voice of the NCC-1701D computer. I've read enough now to know this should be achievable with a modicum of effort via OSS voice cloning tools or training a new model (same difference except "voice cloning" seems to often refer to doing this without training a whole new model?).

Thoughts? Experiences?

I've seen several pages that have led me to conclude this is tenable:

https://github.com/myshell-ai/OpenVoice

https://github.com/domesticatedviking/TextyMcSpeechy

https://github.com/mezbaul-h/june

https://www.home-assistant.io/voice_control/voice_remote_local_assistant/

https://heywillow.io/hardware/#esp32-s3-box-lite

I'm not sure if this would be better for ~music, ~tech, or what, but after messing around with Udio for a bit, I made some stuff I liked and wanted to get folks' thoughts. Imo, it's incredible to be able to get music from a text prompt - it means I, as someone who is mostly ignorant to music production, can have my musical idea and actually render that out as music for someone to hear. I can think "damn that would be cool" and then in kind of a fuzzy way, make it happen then and there. Whether it's good, I don't know. That's not up to me, really, but it is the kind of sound I wanted to happen, so I'm left conflicted on how to feel about it. Figured it would be worthwhile to show folks some of it, and see what they think.

I do enjoy synth and metal, so there's a lot of that in these. Feel free to be as critical as you like. If I can apply your criticism I will try to do it, and if you want to see how that works out, I'll share.

1. Cosmoterrestrial
2. A Floyd, Pinkly
3. Empire's Demise, Foretold
4. Metal for Ghosts Bedsheet Edition (the very end of this one is hilariously appropriate)
5. Multi-3DS Drifting

And here's a link to my profile, if you would like to browse. It will update too when I put more up.

They're all instrumental. Lyrical music is less appealing to me in general and Udio's voices do sound kinda weird to me more often than not. The way I made the tracks, I would start with a clip combining some genres/moods, and then add to either end of the clip until I had a complete song. Along the way, I could introduce new elements/transitions by using more text/tweaking various settings and flipping "manual mode" on and off. The results were fuzzy; I didn't always get what I wanted, but I could keep trying until I did, or until I got something that sounded "better". I wrote all the titles after the song was finished. The album art is from a text prompt.

I'm not sure what I think, to be honest. On the one hand, a lot of the creative decision-making wasn't mine. On the other, the song would not be what it is without me making decisions about how it came about and what feelings/moods/genres were focused upon/utilized. I think the best I can say is "use the tool and see whether it's enough to count". To me it feels almost 50/50, like I've "collaborated with my computer" rather than "made music". Does it matter? If the sound is the intended sound, the sound I hoped to make and wanted to share, is that enough to say it is "my music"? Is this perhaps just what it looks like to be a beginner in a different paradigm?

When I used Suno, I had a much more rigid opinion. What it produced, I called "computer spit". Because, all I could actually control was telling it to continue, changing the prompt, and giving it structure/genre tags that felt like a coin flip in terms of effectiveness. I had a really hard time trying to get it to keep/recall melody, and my attempts to guide it along felt more like gambling than deliberate decisions. It also couldn't keep enough in context to make the overall song consistent with respect to instrumentation. It's different with Udio, both because you have a lot of additional tools, and because it feels like those tools work more consistently at making the model do what you want. I still call the results "computer spit" where I've shown them off, but I'm unsure now whether the production has enough of myself in it to be something more. Perhaps not on the same level as something someone produced by playing an instrument, or choosing samples/arranging things in software, but also not quite the same as the computer just rolling along, with me going "thumbs up" or "thumbs down". Maybe these distinctions don't actually matter, but I'd be curious if anyone has thoughts along these lines.

I'm intentionally trying to avoid a discussion about the morality of the thing or what political/social ramifications it has, not because I don't care about that but because I'm in the middle of trying to understand the tool and what its results mean. Would you consider what I've posted here work I could claim as my own, or do you think the computer has enough of a role to say it's not? Is my role in the production large enough? Or perhaps you have a stronger position, that nothing the computer can possibly do in this way counts as original music. Does any of this change that position for you? I ask because I've gone through a lot of opinions myself as I've been following things, and one interesting bit is that I have not gotten any copyright notices when I've uploaded the music to Youtube (I did get notices with Suno's music). As far as I can tell, with what is available to me, this is all original.

And of course, the most important one: Did you like it? Is there something you think would make them better? Do they all suffer from something I'm not seeing/hearing? I'm not an expert technician nor a music producer, so perhaps my ignorant ears are leading me astray. Either way, I've had a ton of fun doing this, and the results to my ear are fun to listen to while I'm doing stuff. I wouldn't call any of it the best music I've ever heard, but I can also think of a lot that is worse. I think what I wonder the most is whether it comes off bland/plain. Most of the folks I show things to are a bit too caught up in being astounded/disturbed to really give me much feedback, so perhaps putting the request in this form will work out a bit better - ya'll have time to think on it.

As always, your time and attention is greatly appreciated

Edit: I should clarify. I am not attempting to be a musician. Hence calling it "computer spit" with anything public, and the lack of any effort to pitch it as something I did only on my own. Rather, I recognize the limit of my own understanding, and felt I'd hit a point where my ignorance of production meant I could not judge the results as well as I'd like. That means it's time to engage some folks because folks out there are likely to know what I do not and see things I can't. From that angle, a lot of the discussion is very interesting, and I'll be responding to those in a bit. But there's no need to argue for doing the work - I recognize that. I'm trying to see past my own horizons with a medium I don't put the work into. I'm a consumer of music, not a creator, so getting some perspective from folks more acquainted with creating and with the technology is really what I'm after in sharing the experience.

Edit again: Thank you all for a very interesting discussion. I had a spare evening/morning and this was a good use of it. For the sake of tying a bow on the whole thing, I'll share my takeaways as succinctly as I can manage.

It seems, at present, and at best, the role these tools can play is of a sort of personal noise generator. The output is not of sufficient interest, quality, complexity, etc., to really be regarded the same as human-produced music, is the overall impression I have been left with. And for other reasons, it may be that the fuzziness of it all is a permanent feature, and thus a permanent constraint on how far toward "authentic" the results can ever get. I was trying to avoid a discussion about my own creativity, the value of doing work, societal ramifications, etc., so I'll work on how to present things better. For what it's worth, this has all been part of what I do creatively - my area of study was philosophy, and the goal of that to my mind has always been "achieving clarity". So I am attempting to achieve clarity with things as they develop, as a hobby sort of interest while I'm busy doing completely different stuff and to better protect my own mind against dumb marketing and hype. So once again, I appreciate you all taking the time, and I wish you all well in all the things you do.

I'm finding it more and more frustrating having to parse the things I'm actually looking for from what feels like a deluge of rubbish. Are there any strategies, extensions, add-ons, etc that people are using to filter results?

Now that there are some specs, development news, and Snapdragon X Elite vs Intel benchmarks from the past couple days to discuss (with the exception noted below), I thought I'd put together a few links for people. I'm curious how people feel about this iteration of technology in an ARM package for development, tinkering, or edge AI applications. And are folks enthused by the possibilities (Windows or otherwise), dislike the price points, or tired of the AI/CoPilot buzz?

Snapdragon Dev Kit for Windows features Qualcomm Snapdragon X Elite Arm SoC for AI PC application development

New Qualcomm Snapdragon X Elite Benchmarks Show It's a Serious Contender

Qualcomm Snapdragon Dev Kit for Windows is a tiny desktop PC with a focus on AI apps

Debian 12 and Linux upstreaming for the Qualcomm Snapdragon X Elite SoC (January, 2024)

Microsoft is already taking orders for both Surface Pro (11ᵗʰ Edition, $999+) and Surface Laptop (7ᵗʰ Edition, $999+)^[1] orders planned for shipping in June. Each can be bought in X Plus or X Elite flavors, though Wikipedia suggests there are several models of the X Elite so I'm curious which flavors we'll see in MS devices.

^[1] Note that I linked to the "business" versions of the MS Store listings because they get straight to the point with a tech overview, etc. The "business" versions are listed for $100 more than the consumer versions.

Sharing this in ~comp, but if there's a better place for this then I'm happy to see it moved to a more suitable location. Thanks!