I like everything that can be work on CPU without the need for GPU I like Linux mint I like Jan AI I like Ollama and I use it on CPU I like your good channel and thanks for the good content.🌹🌹🌹 In fact, I am more interested on STT rather than TTS 😋 So I may be able to make orders and excution using only voice commands.. Yes I like to be a lazy OS user 😴
Thank you for this series of videos, they are very relevant to me as they align with my interests and skill set. (I’m Currently learning to code python and struggling with cryptic linux commands). I did install Mint 22 as per your suggestion. But on bare metal with a 8GB Nvidia card. Struggled with installing the nvidia drivers and cuda, but after much retrying (thankyou timeshift) I got it working. However I realize you are concentrating on CPU only inference. I followed your videos, installed all those apps. I also installed Ollama, LMStudio and ShellGPT. From someone that struggles with cryptic linux commands in the terminal I found ShellGPT very useful.(but it uses an OpenAI API key). Still looking for an AI assistant to help me learn Python, If there is such a thing. You asked for suggestions, so here is my 2c worth. (I'm sure you have already thought of these). - How to use Coqui TTS with other apps (how to pipe output of LLM to Coqui)? Ollama (as you mentioned). LMStudio (or at least mention it as a alternate to Jan) Shell-GPT (or one on the many other alternatives as listed on the Ollama pages) Any coding assistants? (like copilot, plugins for VSCode or for other IDE’s) - Looking forward to more videos from you.
Very good feedback! Thank you so much for taking the time to write this up. I will take a look and see what I can incorporate and come up with. I do want to build an AI Halloween animatronic that uses TTS and an LLM to interact with trick or treaters, so perhaps that's a good opportunity. Regarding Linux cmds, you are not alone, lol. Have you heard of Warp Terminal, that might be an alternative to what you are using now. I did a quick video on Warp a while back. We can look at building an AI coding assistant, that sounds like a fun project as well. Not sure I'll include it in this series or not, but either way it sounds fun! 🙂. Thanks again for your support.
It is awesome to find your series. It feels like awesome way to pinky dip your fingers in AI. I assume this will do on more or less any Linux - will try to play on Ubuntu, hope it will be with same ease. As for what would be very awesome point to explore - can JAN AI or other tool access online data and/or be provided with other documents in doc/odt/pdf formats for analysis and question in regard of them.
Next we need the other direction, Speech to text, and can we use API's to have a bi-directional voice interactions with a LLM? We need to tie the LLM to Coqui; then find a text to speech to go the other direction. Is there anyway to use something like a json file to setup the LLM on boot, like with a prompt and model?
I haven't explored that much, but maybe whisper? huggingface.co/vasista22/whisper-hindi-small#:~:text=This%20model%20is%20a%20fine,the%20Whisper%20fine%2Dtuning%20sprint.
Do you use a different solution for local TTS?
I like everything that can be work on CPU without the need for GPU
I like Linux mint
I like Jan AI
I like Ollama and I use it on CPU
I like your good channel and thanks for the good content.🌹🌹🌹
In fact, I am more interested on STT rather than TTS 😋 So I may be able to make orders and excution using only voice commands.. Yes I like to be a lazy OS user 😴
@HassanAllaham I appreciate the support!! I might try making an AI assistant
excelent! Thank you
You are very welcome. Thank you for checking out the video! 🙂
Thank you for this series of videos, they are very relevant to me as they align with my interests and skill set. (I’m Currently learning to code python and struggling with cryptic linux commands).
I did install Mint 22 as per your suggestion. But on bare metal with a 8GB Nvidia card.
Struggled with installing the nvidia drivers and cuda, but after much retrying (thankyou timeshift) I got it working.
However I realize you are concentrating on CPU only inference.
I followed your videos, installed all those apps. I also installed Ollama, LMStudio and ShellGPT.
From someone that struggles with cryptic linux commands in the terminal I found ShellGPT very useful.(but it uses an OpenAI API key).
Still looking for an AI assistant to help me learn Python, If there is such a thing.
You asked for suggestions, so here is my 2c worth. (I'm sure you have already thought of these).
-
How to use Coqui TTS with other apps (how to pipe output of LLM to Coqui)?
Ollama (as you mentioned).
LMStudio (or at least mention it as a alternate to Jan)
Shell-GPT (or one on the many other alternatives as listed on the Ollama pages)
Any coding assistants? (like copilot, plugins for VSCode or for other IDE’s)
-
Looking forward to more videos from you.
Very good feedback! Thank you so much for taking the time to write this up. I will take a look and see what I can incorporate and come up with. I do want to build an AI Halloween animatronic that uses TTS and an LLM to interact with trick or treaters, so perhaps that's a good opportunity.
Regarding Linux cmds, you are not alone, lol. Have you heard of Warp Terminal, that might be an alternative to what you are using now. I did a quick video on Warp a while back.
We can look at building an AI coding assistant, that sounds like a fun project as well. Not sure I'll include it in this series or not, but either way it sounds fun! 🙂.
Thanks again for your support.
It is awesome to find your series. It feels like awesome way to pinky dip your fingers in AI. I assume this will do on more or less any Linux - will try to play on Ubuntu, hope it will be with same ease. As for what would be very awesome point to explore - can JAN AI or other tool access online data and/or be provided with other documents in doc/odt/pdf formats for analysis and question in regard of them.
Thanks for the great feedback! I'll definitely be exploring Jan or another local frontend having capability to interact with data. 🙂
Bro thanks so much for the shoutout and I may not be using Docker but Podman
Next we need the other direction, Speech to text, and can we use API's to have a bi-directional voice interactions with a LLM? We need to tie the LLM to Coqui; then find a text to speech to go the other direction. Is there anyway to use something like a json file to setup the LLM on boot, like with a prompt and model?
HI
That's amazing! Is it possible to install the coqui via portainer?
Nice. Is there a way to use it with Calibre Ebook reader?
Yo bro will you show your pc specs and can i use your steps to run local llm on old hardware about 5 years old
It's a vm with 8 virtual cpu cores and 16GB ram. You can run it on less though.
It doesn't support Hindi . Is there anything solution for Hindi
I haven't explored that much, but maybe whisper? huggingface.co/vasista22/whisper-hindi-small#:~:text=This%20model%20is%20a%20fine,the%20Whisper%20fine%2Dtuning%20sprint.