GLaDOS Voice Assistant | Software

[@jensonbridge660]

this is actually epic as I am a big Portal fan, Great job

[@PacAnimal]

Awesome project! The online glados voice generator will serve samples immediately if they've already been generated by someone else. Only new ones require waiting.

[@nerdaxic]

Yeah, it is sometimes helpful when the system can fetch an arbitary number ASAP. This is why I dont have decimal in my temperature readings - integers are usually in stock but floating point numbers are not!

[@PacAnimal]

Have you considered splitting them with special cases in the code as you do for certain words? Regex replace number point othernumber with whole numbers and the word "point"?

[@PacAnimal]

I mean playing multiple samples. The temperature outside is... 14... Point... Two. It's what the black mesa transit system and such do, on purpose, to mimic even older automated phone systems that all do this with pregenerated samples.

[@nerdaxic]

@@PacAnimal Good idea. I am already using that somewhere in the software but cant remember where! Probably reading out humidity readings.

[@amirebrahimi1453]

Amazing project ! Keep going bro

[@nerdaxic]

Thanks for your comment! I got some support to the project, so I might just be keeping at it

[@dergus8833]

yo i looked up to see if anyone had made glados as an ai voice, yours is really accurate and sounds great. good work

[@raveyardz]

I wish i could use this on my android phone. I'd love for her to be in my pocket!

[@Manticorn]

Are you still going to update us if you get the head turning function working? I'd love to see it!

[@nerdaxic]

Id love to see it too! Currently Im like on a third side-project, so Id expect the steward platform to take years to pull off... Lets see!

[@kotcraftchannelukraine6118]

Cool, i making GLaDOS in Python too, but i really making GLaDOS in Python with personality cores, complex AI and ability to control whole system (i make video about it in the future because it not finished)

[@nerdaxic]

I wish you well with your project 🙏

[@kotcraftchannelukraine6118]

Thanks, i also learning neural networks, natural language processing, AI text generation and other to make a real GLaDOS in Python (but only for Ubuntu. Maybe Debian)

[@nerdaxic]

@@kotcraftchannelukraine6118 I abandoned RaspberryPi and switched to Ubuntu VM runnin on my Proxmox server. Much better.

[@claytonjannusch6333]

I've got your GLaDOS-TTS python engine running in a container someone made an image for. Basically what I'm trying to do is integrate your TTS engine into HomeAssistant by having GLaDOS-TTS replace "Piper" in the Local Assist Pipeline instead of running the entire GLaDOS voice assistant suite. I'm guessing I would have to write some python to get your project working over the Wyoming protocl, but if there's some easier way I've missed or another way you can help me out, please do.

[@claytonjannusch6333]

Luckily some other good man has made a package for the source TTS by R2D2FISH that interfaces using the Wyoming protocol. Now I just have to do a bunch of certificate work to actually use the voice assistant in HA, but I got what I was after.

[@dennis6135]

Very cool project. Are there any plans for making your own GLAdOS tts engine? No idea how that would work but I can imagine that would make life easier :)

[@dside_ru]

There's one in Larynx. I've only really fed it a few samples, but those sounded alright to me 🙂

[@nerdaxic]

I have planned to do it yes. It requires some AI knowledge that I do not yet have, but I intend to - this being a good practise project.

[@dennis6135]

@@nerdaxic awesome, I will keep following your process!

[@ironmaniacxd5658]

This is so cool its insane... You are very gifted with coding knowledge, now for a totaly stupid question, can this be Installed on windows 10? If so, how? I dont have any knowledge in python or coding but I rly do want glados on my pc xD... All replys are appriciated

[@nerdaxic]

Dont take this personally, but if you have to ask, then installing this on windows is probably no can do. I have not tested it but technically it should be possible to make it work if all the right Python packages are installed.

[@ironmaniacxd5658]

@@nerdaxic Hmmm good to know, and no need to worry, I dont know shit about coding. Anyhow tnx for the quick response and also again big fan of the project. Would be cool to see glados ai on pc/ android in the near future

[@Necronxx666]

Hey, do you have some goods tutorial to recommend on how to train your own TTS voice ? (Not necessary based on glados). Amazing work btw :)

[@nerdaxic]

Thank you! Unfortunately I do not. I'd start searching for tacotron videos. Mine uses that to generate the basic voice data and then HiFiGAN as vocoder to make it sound like GLaDOS. I've had a lot of help from the open source community with the machine learning stuff, still not completely comfortable with it myself. Advice: just start, youll figure it out eventually. Will be amazing journey!

[@mokey33]

whats that wallpaper in the background its dope

[@nerdaxic]

You mean the video thats running in the background? Its an asset extracted from Portal video game.

[@polyprint3d894]

is there anyw way to make it say german things

[@nerdaxic]

Well yes, but actually no. Would need to rewrite the software to generate the german responses and have a new TTS engine that would then speak that out. Unfortunately GLaDOS is an english character, and I dont think there is training data in the game to train the german model.

[@FieryRobinStudios]

I'm trying to make it rn lol it's hard

[@jamiemitchell3532]

1 question im new to all of this but how would i just get her to run on a laptop

[@nerdaxic]

Should work fine on laptop running Ubuntu

[@dside_ru]

Okay, KZbin keeps nuking my comments, I don't even know what to do anymore.

[@dside_ru]

Does it not like the word Rhasspy?

[@dside_ru]

Okay, seems like I've somehow worked around the pain point. Basically, it's a collection of different components, mostly local, for voice assistants. Hotword > boundary detection > STT > intent recognition > intent handling > TTS.

[@nerdaxic]

Curious. I get the list of rejected or "to be checked" comments, section was empty. Thanks for the keywords, and comment about Larynx, I shall look into these when I got time. Would be awesome to get the voice assistant actually working like an AI voice assistant should instead of being a huge bodge :D

[@dside_ru]

@@nerdaxic thanks! I've heard about similar issues with the comments on LTT's WAN Show, where they were being accused of censorship even, but didn't even get to see the comments 🤔 The biggest discovery and tradeoff with Rhasspy for me was the need to use a closed grammar, where the speech recognizer looks for specific sentences rather than transcribing speech in general. This had a weird side effect where it was recognizing *something* even in random mumbling, until I added in an intent with no handler and a bunch of sentences that would, for lack of handling, be ignored. But for local-only operation I think it's worth it!

[@JohnHELPME]

mind if i ask how do you code this like what goes where just wondering EDIT: also wondering is it possible to hook this up to sometimes like chatgpt to generate responses?

[@nerdaxic]

Check my GitHub. But yes, ChatGPT should be easy to integrate.