#mycroft — Public Fediverse posts

https://www.europesays.com/at/32318/ „Krone“-Serien-Kritik – „Young Sherlock“: Klug, schnell und schlagkräftig #AndrewLanes #ArthurConanDoyles #AT #Austria #chinesischen #ColinFirths #CordeliaHolmes #DónalFinn #Entertainment #FiennesTiffin #GulunShou‘an #GuyRitchie #Hodge #JamesMoriarty #JudeLaw #MaxIrons #Moriarty #Mycroft #NataschaMcElhone #Österreich #RobertDowneyJr #Serien #Series #Sherlock #SherlockHolmes #Unterhaltung #ZineTseng

Over the summer our 6 year old asked me to read to him from The Official Ubuntu Book (which I co-authored). On the page about the #Mycroft assistant device, I mentioned I had one. He was entranced.

Well, yes, I do. A Mark 1. It's been in my closet since I got it. It might be too old to run #OVOS. Or the hardware is dodgy. Maybe I should consider #HiveMind 🤔 Hey, someone upgraded it to use a RPi4.

What kind of hole am I digging for myself here?

My test bed Mycroft / OpenVoiceOS installation is coming alongnicely. It's frustrating that MycroftAI's excellent online portal for managing skills and settings disappeared with the demise of the company. But I'm encouraged by the onscreen developments OVOS has inherited from Mycroft II.

#MycroftAI #Mycroft #OVOS #OpenVoiceOS
#opensource

Curious error installing #mycroft #ovos skills, several packages are seeking a json file, but using a bad line of code:

with open(f"{get_xdg_config_save_path()}/skills/{self.skill_id}/settings.json" "w") as f:

is clearly missing a comma, so Python dutifully seeks 'jsonw', but considering there are (so far) two skills that fail because of this, I'm puzzled as to where to look to fix and/or report the problem.

#funwithopensource

My new puzzle is revisiting an old puzzle - just before they collapsed, I very nearly had #MyCroft in a useful form, but then it all fell apart and was feared lost.

But now there's an active fork: https://openvoiceos.github.io with a flashy GUI and looks very promising, except that #OVOS won't see my HP laptop microphone (UbuntuStudio audio is pipewire, so that could be an issue, but it should emulate ALSA or pulseaudio!)

The newly announced #RaspberryPi5 looks really cool, but I have to admit that I'm a little disappointed the form factor is going to change enough that I probably won't be able to swap out the #Pi4 in my #Mycroft #Mark2 easily. Upgrading the #SSD has been really helpful, but faster RAM, faster USB, and a just straight up faster system would have been really a welcome change. I guess I'll just have to stick to my plan of offloading some services from the device to see if it's enough.

This is one of the reasons I've been a big supporter of #Mycroft in the past, and I'm still supporting #OVOS and #NeonAI. I think a voice driven UI (VUI) is important, but the biggest players in that market are Google and Amazon, and their motivation isn't improving the experience for the people using their devices, it's monetizing anything and everything they can. I'd love to see a new smart speaker (*cough* @PINE64 *cough*) that would support OVOS or NeonAI soon.

https://arstechnica.com/gadgets/2023/09/amazon-wants-to-charge-a-subscription-fee-for-alexa-eventually/

Ooo, just got notified that the new #NeonAI Version 2.0 is available for my #Mycroft #Mark2 That's super exciting. I can't wait to try it out!

https://neon.ai/UpgradeGuideV2.0

I feel like this could be an opportunity for a more open option. Both Google and Amazon have seemingly lost some of their focus on this market because it turns out people aren't just going to ask their voice assistant to buy them something and trust that it's just all going to work out. An open option like #Neon or #OVOS could step in and cover the basic functions. Unfortunately with the demise of #Mycroft, there's no hardware to run it on other than PCs or the #Mark2.

https://www.computerworld.com/article/3703729/whats-going-on-with-google-assistant.html

je me relance dans mes investigations Speech-to-text et Text-to-speech. Bizarrement c'est quelque chose qui revient régulièrement. Est ce que des gens savent ce qu'est devenu #DeepSpeech et si #MozillaVoice est toujours maintenu?
De mon côté je me base sur #mycroft mais j'ai le sentiment que le projet est un peu à l'arrêt mais je me trompe peut-être.

It's good that other people are also bringing up the elephant in the room: why do you need to pay money for one more electronic gadget that listens to you 24/7, when voice assistants aren't supposed to be rocket science in 2023 anymore? https://news.ycombinator.com/item?id=35857631

I wrote two articles on how to build custom #VoiceAssistants using just a Raspberry Pi and a microphone, one in 2019 https://blog.platypush.tech/article/Build-your-customizable-voice-assistant-with-Platypush and one in 2020 https://blog.platypush.tech/article/Build-custom-voice-assistants.
It's definitely doable and I still have my own custom assistants in the house. However, I had to get around with a #Snowboy model for hotword detection (and Snowboy is now basically abandoned), Mozilla #DeepSpeech model for speech-to-text (and that's quite heavy), and #Mycroft's mimic3 text-to-speech model (and Mycroft is now basically bankrupt). Then writing the integration is relatively easy - I used #Platypush, but it can definitely be done with Home Assistant and OpenHAB too.

Compared to 3-4 years ago, I think we're now in a state where the content is no longer the issue (just plug into a LLM, and all of your text requests will get an answer), nor integrations are a problem (just write a Platypush event hook on speech detected, and you can connect it to everything, no need for "Works with Google/Alexa" labels). Text-to-speech synthesis has also become cheap and ubiquitous.

But the hotword detection and speech-to-text models are still IMHO the bottleneck. Hotword detection is a field where you need a very small and lightweight model that only detects a specific word or phrase in a very reliable way. Snowboy was an amazing FOSS project - which also came with this cool idea of "crowd-funded models", where in order to download a model for a certain hotword you were first supposed to provide three audio tracks where you say that word in order to improve the model. But it's now discontinued because it cost the volunteers too much to run the infra.

And Mozilla DeepSpeech is a relatively good choice for general-purpose speech-to-text, but it's heavy (it takes 100% of the CPU when it runs on a Raspberry Pi) and it's mostly optimized for English - even support for other Western languages is patchy. OpenAI's recent Whisper model seems like a solid alternative, but it's also plagued by the 100% CPU issue - also, I no longer trust anything that comes from OpenAI, no matter how noble some of their efforts may look.

If there are other open-source alternatives that solve these problems, I'd be very happy to learn about them. Once these blockers are removed, there should be really no reason for anyone to feed their audio streams to Google or Amazon.

In the meantime, I'm planning to spend some time playing with some self-hosted LLM model to see if I can replace the Google Assistant library on the last Raspberry Pi that runs it in my home.

It's good that other people are also bringing up the elephant in the room: why do you need to pay money for one more electronic gadget that listens to you 24/7, when voice assistants aren't supposed to be rocket science in 2023 anymore? https://news.ycombinator.com/item?id=35857631

I wrote two articles on how to build custom #VoiceAssistants using just a Raspberry Pi and a microphone, one in 2019 https://blog.platypush.tech/article/Build-your-customizable-voice-assistant-with-Platypush and one in 2020 https://blog.platypush.tech/article/Build-custom-voice-assistants.
It's definitely doable and I still have my own custom assistants in the house. However, I had to get around with a #Snowboy model for hotword detection (and Snowboy is now basically abandoned), Mozilla #DeepSpeech model for speech-to-text (and that's quite heavy), and #Mycroft's mimic3 text-to-speech model (and Mycroft is now basically bankrupt). Then writing the integration is relatively easy - I used #Platypush, but it can definitely be done with Home Assistant and OpenHAB too.

Compared to 3-4 years ago, I think we're now in a state where the content is no longer the issue (just plug into a LLM, and all of your text requests will get an answer), nor integrations are a problem (just write a Platypush event hook on speech detected, and you can connect it to everything, no need for "Works with Google/Alexa" labels). Text-to-speech synthesis has also become cheap and ubiquitous.

But the hotword detection and speech-to-text models are still IMHO the bottleneck. Hotword detection is a field where you need a very small and lightweight model that only detects a specific word or phrase in a very reliable way. Snowboy was an amazing FOSS project - which also came with this cool idea of "crowd-funded models", where in order to download a model for a certain hotword you were first supposed to provide three audio tracks where you say that word in order to improve the model. But it's now discontinued because it cost the volunteers too much to run the infra.

And Mozilla DeepSpeech is a relatively good choice for general-purpose speech-to-text, but it's heavy (it takes 100% of the CPU when it runs on a Raspberry Pi) and it's mostly optimized for English - even support for other Western languages is patchy. OpenAI's recent Whisper model seems like a solid alternative, but it's also plagued by the 100% CPU issue - also, I no longer trust anything that comes from OpenAI, no matter how noble some of their efforts may look.

If there are other open-source alternatives that solve these problems, I'd be very happy to learn about them. Once these blockers are removed, there should be really no reason for anyone to feed their audio streams to Google or Amazon.

In the meantime, I'm planning to spend some time playing with some self-hosted LLM model to see if I can replace the Google Assistant library on the last Raspberry Pi that runs it in my home.

It's good that other people are also bringing up the elephant in the room: why do you need to pay money for one more electronic gadget that listens to you 24/7, when voice assistants aren't supposed to be rocket science in 2023 anymore? https://news.ycombinator.com/item?id=35857631

I wrote two articles on how to build custom #VoiceAssistants using just a Raspberry Pi and a microphone, one in 2019 https://blog.platypush.tech/article/Build-your-customizable-voice-assistant-with-Platypush and one in 2020 https://blog.platypush.tech/article/Build-custom-voice-assistants.
It's definitely doable and I still have my own custom assistants in the house. However, I had to get around with a #Snowboy model for hotword detection (and Snowboy is now basically abandoned), Mozilla #DeepSpeech model for speech-to-text (and that's quite heavy), and #Mycroft's mimic3 text-to-speech model (and Mycroft is now basically bankrupt). Then writing the integration is relatively easy - I used #Platypush, but it can definitely be done with Home Assistant and OpenHAB too.

Compared to 3-4 years ago, I think we're now in a state where the content is no longer the issue (just plug into a LLM, and all of your text requests will get an answer), nor integrations are a problem (just write a Platypush event hook on speech detected, and you can connect it to everything, no need for "Works with Google/Alexa" labels). Text-to-speech synthesis has also become cheap and ubiquitous.

But the hotword detection and speech-to-text models are still IMHO the bottleneck. Hotword detection is a field where you need a very small and lightweight model that only detects a specific word or phrase in a very reliable way. Snowboy was an amazing FOSS project - which also came with this cool idea of "crowd-funded models", where in order to download a model for a certain hotword you were first supposed to provide three audio tracks where you say that word in order to improve the model. But it's now discontinued because it cost the volunteers too much to run the infra.

And Mozilla DeepSpeech is a relatively good choice for general-purpose speech-to-text, but it's heavy (it takes 100% of the CPU when it runs on a Raspberry Pi) and it's mostly optimized for English - even support for other Western languages is patchy. OpenAI's recent Whisper model seems like a solid alternative, but it's also plagued by the 100% CPU issue - also, I no longer trust anything that comes from OpenAI, no matter how noble some of their efforts may look.

If there are other open-source alternatives that solve these problems, I'd be very happy to learn about them. Once these blockers are removed, there should be really no reason for anyone to feed their audio streams to Google or Amazon.

In the meantime, I'm planning to spend some time playing with some self-hosted LLM model to see if I can replace the Google Assistant library on the last Raspberry Pi that runs it in my home.

It's good that other people are also bringing up the elephant in the room: why do you need to pay money for one more electronic gadget that listens to you 24/7, when voice assistants aren't supposed to be rocket science in 2023 anymore? https://news.ycombinator.com/item?id=35857631

I wrote two articles on how to build custom #VoiceAssistants using just a Raspberry Pi and a microphone, one in 2019 https://blog.platypush.tech/article/Build-your-customizable-voice-assistant-with-Platypush and one in 2020 https://blog.platypush.tech/article/Build-custom-voice-assistants.
It's definitely doable and I still have my own custom assistants in the house. However, I had to get around with a #Snowboy model for hotword detection (and Snowboy is now basically abandoned), Mozilla #DeepSpeech model for speech-to-text (and that's quite heavy), and #Mycroft's mimic3 text-to-speech model (and Mycroft is now basically bankrupt). Then writing the integration is relatively easy - I used #Platypush, but it can definitely be done with Home Assistant and OpenHAB too.

Compared to 3-4 years ago, I think we're now in a state where the content is no longer the issue (just plug into a LLM, and all of your text requests will get an answer), nor integrations are a problem (just write a Platypush event hook on speech detected, and you can connect it to everything, no need for "Works with Google/Alexa" labels). Text-to-speech synthesis has also become cheap and ubiquitous.

But the hotword detection and speech-to-text models are still IMHO the bottleneck. Hotword detection is a field where you need a very small and lightweight model that only detects a specific word or phrase in a very reliable way. Snowboy was an amazing FOSS project - which also came with this cool idea of "crowd-funded models", where in order to download a model for a certain hotword you were first supposed to provide three audio tracks where you say that word in order to improve the model. But it's now discontinued because it cost the volunteers too much to run the infra.

And Mozilla DeepSpeech is a relatively good choice for general-purpose speech-to-text, but it's heavy (it takes 100% of the CPU when it runs on a Raspberry Pi) and it's mostly optimized for English - even support for other Western languages is patchy. OpenAI's recent Whisper model seems like a solid alternative, but it's also plagued by the 100% CPU issue - also, I no longer trust anything that comes from OpenAI, no matter how noble some of their efforts may look.

If there are other open-source alternatives that solve these problems, I'd be very happy to learn about them. Once these blockers are removed, there should be really no reason for anyone to feed their audio streams to Google or Amazon.

In the meantime, I'm planning to spend some time playing with some self-hosted LLM model to see if I can replace the Google Assistant library on the last Raspberry Pi that runs it in my home.

It's good that other people are also bringing up the elephant in the room: why do you need to pay money for one more electronic gadget that listens to you 24/7, when voice assistants aren't supposed to be rocket science in 2023 anymore? https://news.ycombinator.com/item?id=35857631

I wrote two articles on how to build custom #VoiceAssistants using just a Raspberry Pi and a microphone, one in 2019 https://blog.platypush.tech/article/Build-your-customizable-voice-assistant-with-Platypush and one in 2020 https://blog.platypush.tech/article/Build-custom-voice-assistants.
It's definitely doable and I still have my own custom assistants in the house. However, I had to get around with a #Snowboy model for hotword detection (and Snowboy is now basically abandoned), Mozilla #DeepSpeech model for speech-to-text (and that's quite heavy), and #Mycroft's mimic3 text-to-speech model (and Mycroft is now basically bankrupt). Then writing the integration is relatively easy - I used #Platypush, but it can definitely be done with Home Assistant and OpenHAB too.

Compared to 3-4 years ago, I think we're now in a state where the content is no longer the issue (just plug into a LLM, and all of your text requests will get an answer), nor integrations are a problem (just write a Platypush event hook on speech detected, and you can connect it to everything, no need for "Works with Google/Alexa" labels). Text-to-speech synthesis has also become cheap and ubiquitous.

But the hotword detection and speech-to-text models are still IMHO the bottleneck. Hotword detection is a field where you need a very small and lightweight model that only detects a specific word or phrase in a very reliable way. Snowboy was an amazing FOSS project - which also came with this cool idea of "crowd-funded models", where in order to download a model for a certain hotword you were first supposed to provide three audio tracks where you say that word in order to improve the model. But it's now discontinued because it cost the volunteers too much to run the infra.

And Mozilla DeepSpeech is a relatively good choice for general-purpose speech-to-text, but it's heavy (it takes 100% of the CPU when it runs on a Raspberry Pi) and it's mostly optimized for English - even support for other Western languages is patchy. OpenAI's recent Whisper model seems like a solid alternative, but it's also plagued by the 100% CPU issue - also, I no longer trust anything that comes from OpenAI, no matter how noble some of their efforts may look.

If there are other open-source alternatives that solve these problems, I'd be very happy to learn about them. Once these blockers are removed, there should be really no reason for anyone to feed their audio streams to Google or Amazon.

In the meantime, I'm planning to spend some time playing with some self-hosted LLM model to see if I can replace the Google Assistant library on the last Raspberry Pi that runs it in my home.

With the help of @NeonClary, I just upgraded my #Mark2 to an SSD, which has greatly increased the responsiveness. The touch screen is noticeably faster than it was with the USB drive, and it responds to voice commands faster by an order of magnitude. Absolutely worth it if you've got a #Mycroft Mark2 that you've upgraded to Neon. :neon_ai: :mark2:

Not trying to shill for #Mycroft, but right now they have for sale what could possibly be the last run of #Mark2 devices. They've already started shutting down operations, so I doubt they're going to be making more. They're a little on the pricy side, but you can still install Neon AI (https://neonassist.com/neonaidownloads) and keep these devices fully functional with an #OpenSource virtual assistant. If that's not your bag, they have surplus Pis available, and those are hard to find.

https://mycroft.ai/shop/

Support OpenVoiceOS as Nonprofit Association

https://gofund.me/4357e19b

#mycroftai #mycroft #openvoiceos #ovos

#Mycroft ‘s privacy-first, #crowdfunded smart speaker will ship, but not to backers | Ars Technica

Echo alternative's #privacy focus is worth emulating, despite Mycroft's failure.

https://arstechnica.com/gadgets/2023/02/mycrofts-privacy-first-crowdfunded-smart-speaker-will-ship-but-not-to-backers/#p3

A new release of the #Mycroft software is now available. If you own a Mark 2 and you haven't moved to #OVOS or #NeonOS, you should probably update sooner rather than later.

* Add STT fallback to communicate directly with a free demo Google STT service.

* Add config switch that allows Mycroft to start up without valid pairing information.

* Add direct call to Weather API as fallback mechanism.

They're removing dependence on their backend, which is good. No idea how much longer it'll be there.

I'm sad about the Mycroft announcement, but on the other hand, it's already essentially obsolete.

With advances like the Whisper voice recognition system, the Deepspeech TTS project, and now the OpenAssistant project using language models, we're on the cusp of a much more fluid, complete assistant that is privacy respecting, realistic, and much more powerful.

#DeepSpeech #OpenAssistant #Mycroft #Whisper #AI #FLOSS

Ce week-end, je me suis amusé sur mon poste avec #mycroft et #mimic3 (text to speech) en français. Bon, on va pas se mentir, ce n'est pas parfait mais j'arrive à interroger wikipedia et lancer des musiques sur spotify. La prochaine étape est de changer le "wakeword" puis peut etre la partie "speech to text" qui est encore celui de google indirectement. Le plus bizarre / marrant est la lecture de mot anglais en français 🥖

@fsf

I don't use any proprietary #smartSpeaker garbage, but there isn't really any #foss alternative that you can easily #selfhost a backend for. #Mycroft is probably closest as far as feature parity, but they're backend, while #openSource, is not very easy to implement. It's something they are "working on."

I know #rhasspy is a thing, but it isn't very good at handling non completely predefined commands.

@JoeRess I've been wanting to build smart speakers with SBCs because of their form factor. The problem I've run into (other than availability) is just getting audio peripherals that work without standing on top of any device. There's a lot of support from #Mycroft for Pi based devices since the #MarkI and #MarkII are both Pi based devices. A mini PC would work fine for this as well, but I don't want USB devices hanging off it everywhere. The "appliance" device is more what I'm looking for.

I'm looking through the #Mimic3 voices for #MyCroft on the mycroft.ai website, and it's vast. There's so many languages and accents to find. What I find odd, though, is that German language speech synthesizer has emotions on it. The system can speak in a drunken, sleepy, or amused tone. We have truly come a long way with this technology. (BTW, no one needs to tell me about the origins of these samples.) #TextToSpeech #TTS

@mahmoudajawad Hardware wise, they've been abysmally slow. Software wise, they've been OK. They've recently put out #Mimic3, which is really good actually. It's separate from #Mycroft, so it can be used for pretty much anything, and its quality is up there with any TTS solution I've heard running on a local machine.

Ooo, installed #Mimic3 in my #WSL #Ubuntu, and configured my new #Mycroft instance to use it. Works beautifully. I wonder if I can get all this running on the #SteamDeck when it comes in. It's #Arch after all, so I don't see why it shouldn't work.

A fun and slightly weird introductory video of #Mimic3 by #Mycroft. Watched it last night with my wife. Her reaction was, "WTF was that?" Not inappropriate, but still funny.

https://yewtu.be/watch?v=KWw9LKSfAq0

@pixelherodev I haven't tried this personally, so grain of salt here, but it's my understanding that the entire stack CAN run offline now. The #Selene part is still the biggest PITA as they don't have easy configuration to use it and it's setup is a pretty big pain. It's also not a turn key thing to switch #Mycroft installations from the default cloud version to the locally hosted version. #Mimic3 is by far the easiest to use. Despite all that, it is an option to run everything fully on prem.

#Mycroft's #Mimic3 is now live and ready for use. If you've got a project that requires TTS, it could be a very good option for you. It respects your privacy, runs fully offline, and it's open source. On top of all that, there's hundreds of voices to choose from and speaks dozens of languages.

https://mycroft.ai/blog/introducing-mimic-3/

@blackfire Nothing 3rd party, and I'm being more than a little hyperbolic, but there are instructions on the Github page:

https://github.com/MycroftAI/selene-backend

Theoretically with the #Selene backend on prem, any #Mycroft instances pointed to that backend, and a local #Mimic3 install, you can run the whole stack privately and offline. I've never done the whole stack, and Mimic3 is still so new that you have to request the files manually, but I'd love to see this working.

@doomsdayrs @popey

#Mimic3 is finally available for beta testing. It's not a "fully open" test and they send you a link with a generic user/pass combo after you register your email. I assume this is to limit bot traffic and gauge interest. Still looking forward to the #Mycroft #Mark2 in September, and more than likely I'll be using the new @popey voice. I'm still using the original ap voice on my Mark1, and it wouldn't feel natural to use anything else.

https://mycroft.ai/blog/mimic-3-preview/

Hackaday Links: October 18, 2020 - Remember subliminal advertising? The idea was that a movie theater operator would splice a single ... - https://hackaday.com/2020/10/18/hackaday-links-october-18-2020/ #subliminaladvertising #supermassiveblackhole #assistivedriving #hackadaycolumns #radioastronomy #hackadaylinks #hydrogenline #patenttroll #nobelprize #exploit #insight #mycroft #tesla #mars #mole #hp3 #sdr

#PlasmaBigscreen on your #postmarketOS device, soon this can be a reality! I have a #Pine64 A64LTS outputting it to my TV.

Still some work to do, like getting remote control working, but it's getting there. And yes, this is with #MyCroft integration!

https://gitlab.com/postmarketOS/pmaports/-/merge_requests/1552

OPARP Telepresence Robot - [Erik Knutsson] is stuck inside with a bunch of robot parts, and we know what lies down that path. H... - https://hackaday.com/2020/08/31/oparp-telepresence-robot/ #2020hackadayprize #facialrecognition #personalassistant #telepresencerobot #thehackadayprize #objectdetection #telepresence #robotshacks #tensorflow #follower #mycroft #mobile #robot

I've been packaging #mycroft for Alpine Linux lately and finally got it to talk to me after I asked it a question today. Hopefully we can run it on our phones and maybe even TV's (#plasmabigscreen) soon!

Chatterbox - a DIY smart speaker for kids.

Powered by Mycroft (an open source voice assistant, that can be installed on Linux).

Chatterbox is the world's first educational build-it-yourself, program-it-yourself smart speaker kit for kids putting privacy first.

Think of it like an Alexa or Siri, but focused on privacy that kids can build / program to bring their creativity to life!

== > https://www.kickstarter.com/projects/hellochatterbox/chatterbox-the-smart-speaker-that-kids-build-and-p?ref=project_build
_
#Chatterbox #Linux #Mycroft #opensource #Alexa #Siri #AI #voiceassistant #kids

Today, I will be co-hosting on #FLOSSWeekly in about 10 minutes (on #leolaporte's #twittv) to interview the CEO of #Mycroft AI! I am so excited for this episode! Don't miss it! - https://twit.tv/live