thats what im doing, im currently in process of making a package similar to memgpt but with more agents + out of the box support for agents + some custom agents of my own and alot of customization made easy for developers
@@aniksamiurrahman6365 you can't train a local llm, and why would u need to train it locally anyways?, updating is easy it's a drop and replace most the time
Man you are awesome! I have learned a lot of linux scripting from you, but mostly, you make me want to learn more every time, you make my child curiosity kick back in. I love you, nohomo
just use "whisper tiny or base models" , they are much much accurate even with noise in surroundings. also use "porcupine" for wake word detection (very fast and accurate) , instead of typing command every time
Thanks! I used your explanation to make myself a script binded to a hotkey, i implemented the visualizer cava to add a bit of charm to the voice. Pretty cool!
Nice vid like everytime but i have a suggestion,please clearly mention the all the tools you used to get the thing working in the starting,it took me quite some time to figure out that you were using vosk for voice to text.
You can integrate this with optional voice to text, so you can make a deamon and a cli to communicate with it, so you setup one keybind to voice to text, and another for a rofi prompt or something like that
If you include a voice auth system that authorizes a given user for commands and generates a token, then you can have user privileges. its basically a user-detection system, and then your command libraries are more safe. i'd also, for now, use the "windows/macosx/penguin" command key be the "push-to-talk" key, and then you can move that to a peripheral. star trek comms badge, anyone? Brilliant work.
WOW DUDE, THAT IS INSANE! When I've heard your enthusiasm I was expecting to be disappointed, but daaaamn. Thank you for your videos, it's content dense and very helpful! Greetings from Serbia!
Absolutely fucking wonderful. I have so many use cases for this kind of library and script. There's so many little things you could abstract away into commands, tons of building blocks that could be made, not to mention the glory of a truly personalized AI companion. So cool.
Great idea to combine these tools and good tool recommendations in the first place. As others here suggested I might attempt this with Mixtral as well^^ I might need a better PC though
this is something I've been wanting to do for a while! I'm nowhere near as experienced but what I wanted to accomplish was to have a similar outcome, but I wanted the AI to be a locally hosted one. Bonus points for if it automatically saves data from my queries and can trains itself off those so it has a sort of memory. Not even sure where to start on that last part but your video has revitalized my spark to continue thinking about it!
Great tool I didn't know about. I'm thinking of building a sort of Jarvis for an elderly family member. She's getting to the point where any kind of automation would be a quality of life improvement, and to be honest, she's scared of computers - will, not *of* them, but of the interface. She knows how to use a smart phone but not a keyboard and mouse :) Thanks for pointing this out.
Me and you are the same. I love futuristic things and currently I am trying to customize hyprland to just be like that. this is amazing keep up the good work. also you move so fast, like you type and click so fast that I was surpised one second your typing something and 2 milli seconds later you're in the terminal and have going through 15 directory's.
OMG that's perfect! This is what I want my computers to do and it's so obvious, I don't understand why LLM's hasn't been utilized like this from the get go, but now I know how, thanks BugsWriter!
I love this type of stuff!! My wife is going to hate me this weekend. Also.. At least Debian systems can just work after the install, and don't take half a day to finish setting up. Love the videos!! Can't wait to see what else you do with piper.
Linux user asks ChatGPT why people still use Windows, gets upset when AI tells him why people still use windows. Tbh I thought about this the moment LLMs started becoming really good. Getting a local LLM and integrating it deeply into the system to assist with all sorts of things would be amazing. I'm surprised that I haven't seen anyone try to package these tools together to make them easy to set up and use on Linux but I think the limitless potential is amazing. Also it'd be really cool to sync it to a phone and have basically your own jarvis like assistant always with you.
Cool video, as always The possibilities are endless! The only drawbacks I'm seeing is having to go through, and be dependent on OpenAi, for all the data has to pass through it (delay and privacy) and the fact that we're still limited by the model's barriers and character limit. It would be great if you we really could create are own ML/AGI models tailored to us/our needs, that would evolve, learn, but locally, maintaining our privacy.
it's very possible! FastAI is the easiest to pickup python library for deploying local models that I know about. It's not gonna be as generally useful as GPT though. It's a pipe dream for me too, but seems to rapidly be getting more approachable
Quite interesting. The only problem, I believe, is related to the delay that occurs. There is a method that uses the graphics card to change the voice, like w-okada/voice-changer. I don't know if you've heard of it, but one thing that could be added is something like a real-time translator.
Very nicely done. Marrying on-the-fly content generation with speech recognition and synthesis could lead to some very fascinating user interface designs. I feel like we're going to quietly see bes t practices for "intelligent audio" slowly emerge and no one is going to realize that there's a new "normal" in the world until it's done.
@@xDeFc0nx I'm not discrediting the use of this. There have been attempts at accessibility in the form of screen readers, different controllers, etc. and I definitely can see this being integrated into more operating systems in the near future. Heard some buzz about samsung integrating an ai feature into android for example, it's just that personally I'm well versed with a keyboard and see myself being more efficient with it in most day to day applications (for now that is)
I wonder if there’s a way to give it access to what you’re seeing so if you’re programming in vim you could ask project questions or if there’s a complication or runtime error ask what it thinks of it. That would be cool
I can see this being a really powerful plugin to Nushell, which can handle data output in a much more tabular format, also, it would be powerful with agents embedded with certain tools. Do you have any open source intentions for this kind of work flow?
what im planning on doing is having multiple spearkers (with mic) throughout my house which i can connect to with bluetooth, i can have a small rasberrypi connected to them via bluetooth, and rest you already guessed it probably 💀
The biggest problem i see is, ChatGPT costs over 30 cents per query. OpenAI is burning through enormous amounts of money, and they can't do it forever. They could go pay-as-you-go at any time, which takes it out of hobby territory. Maybe you could switch to something like Llama on a local server?
I actually made the same thing and made it as executable and just that i used gemini instead of chat gpt, Name it chat cli, and offer used to use any LLM they want, using local language models will cost performance so your wish there
Thanks Tobias Frisch / TheJackiMonster
for letting me know about this tool.
- thejackimonster.de/blog
for nixos, it's packaged as `piper-tts` and voices as `services.wyoming.piper.servers..voice`
i am surprised that it's not in aur.. or is it?
I am not prepared for "keyboard is bloat"
Honey, its 4:30 am, time for your daily bugswriter content
He posting is american schedule
Now imagine using this with a local LLM and embeddings for unlimited memory its going to be huge.
thats what im doing, im currently in process of making a package similar to memgpt but with more agents + out of the box support for agents + some custom agents of my own and alot of customization made easy for developers
But how do you train a local LLM? How do you update?
@@aniksamiurrahman6365 you can't train a local llm, and why would u need to train it locally anyways?, updating is easy it's a drop and replace most the time
@@ckpioo good luck ❤
@@ckpioo So its just chatGPT/Bard/Llama under the hood.
10:12 Shots fired at mental outlaw
He just reads news lol🤣🤣🤣
lmaoooo
Jarvis, delete my system
J: thank you for ending my suffering
my heartbeat raises when seeing such great creativity possible in linux
That's huge bro 😮
Now use mistral or mixtral for a local LLM
Llama 2
Dolphin-2.5 is the best
This is among the coolest of things I've seen. Good job and your enthusiasm is a joy to witness.
Holy, I have thought about this for a long time, it's amazing to see someone actually do it. GJ mate
Man you are awesome! I have learned a lot of linux scripting from you, but mostly, you make me want to learn more every time, you make my child curiosity kick back in. I love you, nohomo
This is a really great showing of putting she creativity and several programs together! 🔥🔥🔥
just use "whisper tiny or base models" , they are much much accurate even with noise in surroundings.
also use "porcupine" for wake word detection (very fast and accurate) , instead of typing command every time
The only reason why I will ever use Linux: an AI can jury rig all of the software I need for me because Linux is a hellscape for ADHD
Looking forward to what you come up with. This is a game changer
Thanks! I used your explanation to make myself a script binded to a hotkey, i implemented the visualizer cava to add a bit of charm to the voice. Pretty cool!
Last pieces will be having local LLM agent , and passing/choose files using voice .
ahhahahah automate mental outlaw
mental outlaw just read news 💀💀
What happened
Lol
Aw! I'm so going to play with this. Nice video.
Nice vid like everytime but i have a suggestion,please clearly mention the all the tools you used to get the thing working in the starting,it took me quite some time to figure out that you were using vosk for voice to text.
I'm stuck there too ! Could you elaborate me on the configuration??
You can integrate this with optional voice to text, so you can make a deamon and a cli to communicate with it, so you setup one keybind to voice to text, and another for a rofi prompt or something like that
The daemon maybe is too much but there can be 2 scripts calling these programs with different configurations for the same effect
This is great content! Definitely helpful for my further projects :D
@4:15 "If you are smart enough", sir, im not smart, but i definitely see where you're going
If you include a voice auth system that authorizes a given user for commands and generates a token, then you can have user privileges. its basically a user-detection system, and then your command libraries are more safe. i'd also, for now, use the "windows/macosx/penguin" command key be the "push-to-talk" key, and then you can move that to a peripheral. star trek comms badge, anyone?
Brilliant work.
I like your attitude and enthusiasm, keep up the good work. :)
WOW DUDE, THAT IS INSANE! When I've heard your enthusiasm I was expecting to be disappointed, but daaaamn. Thank you for your videos, it's content dense and very helpful!
Greetings from Serbia!
Absolutely fucking wonderful. I have so many use cases for this kind of library and script. There's so many little things you could abstract away into commands, tons of building blocks that could be made, not to mention the glory of a truly personalized AI companion. So cool.
bro this was awesome, i also have few ideas to work on. Thanks
Great idea to combine these tools and good tool recommendations in the first place. As others here suggested I might attempt this with Mixtral as well^^ I might need a better PC though
this is something I've been wanting to do for a while! I'm nowhere near as experienced but what I wanted to accomplish was to have a similar outcome, but I wanted the AI to be a locally hosted one. Bonus points for if it automatically saves data from my queries and can trains itself off those so it has a sort of memory. Not even sure where to start on that last part but your video has revitalized my spark to continue thinking about it!
Great tool I didn't know about. I'm thinking of building a sort of Jarvis for an elderly family member. She's getting to the point where any kind of automation would be a quality of life improvement, and to be honest, she's scared of computers - will, not *of* them, but of the interface. She knows how to use a smart phone but not a keyboard and mouse :) Thanks for pointing this out.
Me and you are the same. I love futuristic things and currently I am trying to customize hyprland to just be like that. this is amazing keep up the good work. also you move so fast, like you type and click so fast that I was surpised one second your typing something and 2 milli seconds later you're in the terminal and have going through 15 directory's.
OMG that's perfect! This is what I want my computers to do and it's so obvious, I don't understand why LLM's hasn't been utilized like this from the get go, but now I know how, thanks BugsWriter!
Ok, that is really a banger. It gives so many ideas for great quality of life things you can make with it. :D
Glad you like it!
You look good 👍
appreciate all the work you did, very cook. bro I really NEED this wallpaper, ricing my setup now and this wallpaper would fit nice
This is the kind of thing I want to mix with the home automation stuff I thinker in.
This is exactly what i was thinking about but never figured out how to do it because im a noob, thanks!
This is one of the coolest things I have ever seen. Please make some videos based on this with some ideas. I have made some wacky shit so far. :)
Thanks for the video man.
I love this type of stuff!! My wife is going to hate me this weekend. Also..
At least Debian systems can just work after the install, and don't take half a day to finish setting up.
Love the videos!! Can't wait to see what else you do with piper.
Linux user asks ChatGPT why people still use Windows, gets upset when AI tells him why people still use windows. Tbh I thought about this the moment LLMs started becoming really good. Getting a local LLM and integrating it deeply into the system to assist with all sorts of things would be amazing. I'm surprised that I haven't seen anyone try to package these tools together to make them easy to set up and use on Linux but I think the limitless potential is amazing. Also it'd be really cool to sync it to a phone and have basically your own jarvis like assistant always with you.
ios18?
This is awesome. I made something like this using ollama and whisper, but I used a really old voice synthetizer. I will give Piper a try. :)
BRO you are a MENACE, keep it up, you're fire
Do you have any progress regarding this or a github repository?
The "so not good, she was just bullshiting" got me hahaha you are great.
Let's make "A.D.I.T.Y.A" a reality then
Bugs, I am here for the ❤. Thank you and have a great day!
the keyboard sounds are perfect
Cool video, as always
The possibilities are endless! The only drawbacks I'm seeing is having to go through, and be dependent on OpenAi, for all the data has to pass through it (delay and privacy) and the fact that we're still limited by the model's barriers and character limit.
It would be great if you we really could create are own ML/AGI models tailored to us/our needs, that would evolve, learn, but locally, maintaining our privacy.
it's very possible! FastAI is the easiest to pickup python library for deploying local models that I know about. It's not gonna be as generally useful as GPT though. It's a pipe dream for me too, but seems to rapidly be getting more approachable
That’s super cool, bro send commands so quick
Quite interesting. The only problem, I believe, is related to the delay that occurs. There is a method that uses the graphics card to change the voice, like w-okada/voice-changer. I don't know if you've heard of it, but one thing that could be added is something like a real-time translator.
Very nicely done. Marrying on-the-fly content generation with speech recognition and synthesis could lead to some very fascinating user interface designs. I feel like we're going to quietly see bes t practices for "intelligent audio" slowly emerge and no one is going to realize that there's a new "normal" in the world until it's done.
Disable capslock function and macro your inevitable creation to the same key.
Amazing Project!!
Great video and we are very closer to an megaman network future
The fact you can run it locally makes it better than ElevenLabs
Great video!
This is amazing! Thanks!
Use SGPT (cli tool) and use the -s flag to get a shell command output. So you can ask it to do something and actually get a command to run.
I know. I had this idea for very long.
Hey i think the problem is llm’s you need an llm that can break down what you really wanna do
Dude, you are my inspiration for becoming tinkerer..❤
And you deserve much more subs and likes!
What vosk model do you use?
i was expecting either linus's or stallman's voice
i liked your environment. Are you using zsh or something else? Can you explain how to make it look like that?
I have a video on zsh
.NET already has a speech synthesizer class, although Piper has better audio.
BRO THIS IS SICK
Bro ist faster than light when typing xD
This is very cool!
Pretty cool, but where can I find the `voice_to_text` ?
same question
Loved it!
I wish there were more voice models, precisely more popular voices like friday or jarvis
what keyboard are you using?
bloat
I really want a Japanese voice assistant who tries to speak English.
Silero tts also pretty good and you can run it locally
Great content as usual.
THATS FUCKING IMPRESSIVE
Wow, this is super cool.
Pretty cool! I like building tools like this too but the chances I'd actually use this are slim.
This is a game changer for people who can't see, are color blind, or the rest. Many uses.
One use case that I could think of would be in learning a new language.
@@xDeFc0nx I'm not discrediting the use of this. There have been attempts at accessibility in the form of screen readers, different controllers, etc. and I definitely can see this being integrated into more operating systems in the near future. Heard some buzz about samsung integrating an ai feature into android for example, it's just that personally I'm well versed with a keyboard and see myself being more efficient with it in most day to day applications (for now that is)
@@bowiemtl I'm the same, I just use this for cool stuff
Is the thumbnail art sourced from somewhere? I like it a lot
very nice one!
I wonder if there’s a way to give it access to what you’re seeing so if you’re programming in vim you could ask project questions or if there’s a complication or runtime error ask what it thinks of it. That would be cool
I think espeak could work well if you could install a better voice. I just don't know where to find them.
I can see this being a really powerful plugin to Nushell, which can handle data output in a much more tabular format, also, it would be powerful with agents embedded with certain tools. Do you have any open source intentions for this kind of work flow?
waiting for more cool stuff you do
ive been looking for ways to train custom voice models like this.
Do you have a proper installation guide? You should make a bash script and have that be opened and started with a macro or key command.
Nice Vid! May i ask your background source?
i love how happy you are lol
Which VTT are you calling with $(voice_to_text)?
Dude i appreciate the love, but how about a response 👍
What is your terminal font it's super pretty
jetbrains mono
what im planning on doing is having multiple spearkers (with mic) throughout my house which i can connect to with bluetooth, i can have a small rasberrypi connected to them via bluetooth, and rest you already guessed it probably 💀
What I was always looking for :P turn my computer in my personal voice assistance...
So tgpt runs in someone's else computer and piper can run on a pi, so... could this setup be replicated on a smartphone, using termux?
The biggest problem i see is, ChatGPT costs over 30 cents per query. OpenAI is burning through enormous amounts of money, and they can't do it forever. They could go pay-as-you-go at any time, which takes it out of hobby territory. Maybe you could switch to something like Llama on a local server?
It's so impressive how fast you can navigate around your desktop and termial. What DE is that?
current wm and configs please?
now i realize that , i was running linux as windows😅
Jarvis, find this guys house
Hello sir, I am your big fan... Can you tell me where did you learn this linux knowledge and other stuff... In college or self... Please answer
laptop + internet
I actually made the same thing and made it as executable and just that i used gemini instead of chat gpt, Name it chat cli, and offer used to use any LLM they want, using local language models will cost performance so your wish there