วีดีโอ

I am using 2 GB version of the Raspberry Pi

That is a great question. The answer is Yes. Large language models (LLMs) like Gemini are trained on massive data. While you may not be able to train AI from scratch, you have the option to fine-tune the LLMs. You provide additional data (answers) tailored to your desired questions, and the LLMs adapt.

@@TechMakerAI This is my School Project. After A long search I found this video and Thank you that this works. Now Can you help me more please? I want to train the bot that when I ask the bot a question, it answer me such a format that I trained it. How to do that. please help me

It sounds like you want to fine turn the LLM. Please have a look at this article: www.datacamp.com/tutorial/fine-tuning-large-language-models

@mukagedek4010: The Google Gemini API and Google TTS are currently free to use. The TTS will most likely be free in the future too. However, it is hard to say about the Google Gemini API. Given that the ChatGPT API is not free, Google may charge a fee in the future for Gemini or they may offer a limited version for free. Anyhow, I would say go ahead and build it. There will be some free options available. Even if Google charges a fee, I think it is still worth it. By the way, in this channel, we are also building companion robots. Please subscribe as you will see in future videos.

Uncomment Line 23 by removing "#" and add your Google Gemini API key

Yes, a USB speaker will work too. I am working on a version of the hardware and will have a video about this topic soon. RPi 3B may work but will be slower due to the small memory size. The wake word is "Jack". To end the conversation with AI, you just need to say "That is all". It will continue to listen but will not respond. Just like Siri, it is always checking to see if you are saying the wake word. If you do not like to see the printed output, just comment all the print() statements.

@@TechMakerAI can you help me change the wake work if you can?

@@TechMakerAI can you make a video on usb speaker sonn cz I want to make this and present it in an exhibition after 2 weeks....I will be very grateful!

Just do a search and replace of all the "jack" with any name your like.

@m9playz239: You can use the current version of the code with a USB speaker. It should work too.

Thank you, will do!

Glad to hear that!

Yes, you can use Gemma model with Ollama. I have an intro video about Google Gemma model, you can find it here th-cam.com/video/QBrjVtMVU3U/w-d-xo.html

Thank you for your feedback! I am aware of the communication issue with Google Gemini API. It can cause crash sometimes. I have seen less issues with the OpenAI ChatGPT API but that is not free. I made some changes to the code on line 259 in this version. When you say "That is all", it should stop interacting with Gemini API but it is still printing out what you are saying. You can comment out all the print() since you do not need it to print transcript when using RPi (It is there only for debugging purpose) or you can move lines 259 to 263 after line 270 (if "jack" in text.lower() and slang == "en-EN":). Note that this code was written so that users can interact with AI anytime they want. They can wake up AI and put it in sleeping mode. This voice assistant should be on all the time like Siri on your phone. If you stop it from listening, then how do you wake up it next time?

@@TechMakerAI Got it. I have follow up questions. If I do not say "That is all", how many minutes until it times out which will require the wake word? Which line can I modify this? Also, I already have an OpenAI API Key i paid for. How can I switch over to Open AI???

@xpectz: For your first question regarding how many times it takes for it to require the wake word, I think you can add an integer variable to count the times and then set the AI to sleeping mode after it reaches a threshold. For the OpenAI version, please see this video: th-cam.com/video/kIKgl3jDilY/w-d-xo.html

@@TechMakerAI How does it work now? When I stop talking to it, it goes to sleeping mode. What triggers it? I have not timed it. Will check out the video, thanks.

If you add an integer variable to count the times and then set the AI to sleeping mode after it reaches a threshold, then, it will be triggered by your integer. To wake it up, you just need to say the wake word again.

Please see my reply to your newer question.

Hi, @pixyful7025, I am glad it is working. It is possible to automatically launch this Python program at the startup of RPi. I have a video on this topic coming up very soon. Please subscribe to my channel as you will be notified when it is out.

​@@TechMakerAI I found a little thing that seems to work by tinkering. I wrote it in .bashrc and told the raspberry to log in automatically

@@TechMakerAI I will wait for your video to compare the most optimized 😉

Yes, it will work with a Bluetooth speaker and without LEDs. I have used the Bluetooth speaker before, see this video th-cam.com/video/nHpJaE559r4/w-d-xo.html. But it requires me to run some commands to pair the Raspberry Pi with the Bluetooth speaker each time. So I switched to wired speaker after that.

You can try some free ones like Llama-3 th-cam.com/video/X8rpOSKDg6I/w-d-xo.html

Great question! That can be a topic for a new video. RPi zero 2 W may not have the computing power to run a Gemini-powered voice assistant.

Hi, I had a similar problem with my old USB microphone few months ago. I got a new one recently. It has been working well. Do you have another microphone to try? The code should work on RPi 5.

Thank you!

Have you installed the OpenAI python package as mentioned on the GitHub page?

Yes, it will. The procedure to create Python virtual environment will be the same on Pi zero 2 W.

@ADOPLAZ: Hi, it should "work" on Pi zero, but you may need to use another text-to-speech engine such as gTTS because the audio quality of the Pyttsx3 package is not good on Pi. Please have a look at this video th-cam.com/video/kE95xA5jVWQ/w-d-xo.html to see how you can use gTTS.

@@TechMakerAI thanks bro, It is for my school project

Check your mic and make sure it is working. Also check the speech recognition part, and make sure it is talking to your mic.

you have to start the convo with "hey jack..." say to it while its listening "hey jack, tell me a joke" make sure jack is said clearly

Hi, @cyberdevil657, for business inquiries, you can find my email by going to www.youtube.com/@techmakerai and then clicking "more" under my channel name.

@@TechMakerAI I replied my friend.

Hi, @Yailya01, that is a very good question. I have a video coming up about this topic. Please subscribe to my channel as you will be notified when it is out.

@@TechMakerAI I've sub and bell, waiting for your next content

Yes, it is possible. Meanwhile, you can change the "text" variable in the code and run it again, it will generate an image according to your input.

You’re welcome 😊

Ok. Thanks

Thanks! Glad you liked it.

@Yailya01: Please check out my new video in which I showed how you can change the language: th-cam.com/video/kIKgl3jDilY/w-d-xo.html

@@TechMakerAI I have followed and watched the latest but when I spoke there was no transcript of the speech to text

@@Yailya01 What language option (slang) are you using in the "rec.recognize_google()" function? It works well for the English language when I set the variable "slang" to "en-EN".

You are welcome.

@Skooly-u7w: Yes, you can speak or ask questions in another language. Please check out my new video here in which I demonstrated how to do it: th-cam.com/video/gYFGHyB_VJ0/w-d-xo.html.

@@TechMakerAI I can see what you did there but I am trying to get it speak dialect not native language

@Skooly-u7w: In that case, you will need to find a text-to-speech engine that supports your local dialect. Once you have that, then you can use it with my Python code.

@hugogarcia2681: You will need to install a media server such as Plex on your Pi.

Thank you! Cheers!

Thanks and welcome.

Hi, @areebmianoor4449, Thanks. The large language model can handle multilingual questions "dynamically", but the speech recognition and text-to-speech conversion parts of the code need to know the language first. Otherwise, the result will not be good. If this is something you want to do, I would suggest you try some multilingual TTS and multilingual speech recognization models.

Hi @ZBZBZB417: thanks! Very good question. Yes, it will work on Raspberry Pi. I have been working to improve these delays. Please subscribe to my channel as you will be notified when the new video is out.

Hi, for business inquiries, you can find my email by going to www.youtube.com/@techmakerai and then clicking "more" under my channel name.

Hi, have you installed all the packages for this project? I have a new version of this code which I talked in this video th-cam.com/video/kE95xA5jVWQ/w-d-xo.html. Maybe you can give that a try.

Glad you like them!

Yes, that is the plan. Please subscribe to my channel so that you will be informed once that video is published.

Hi, @user-bm3rg7jf8g, you can find my email by going to www.youtube.com/@techmakerai and then clicking "more" under my channel name.

how did you enable the venv

Please take a look at this document: docs.python.org/3/library/venv.html

But what did you click to appear the “source ~/.venv/bin/activate “ ? Because when i typed that, there was an error as far as I remember.

Hi, I have a new video about this topic. It will be online very soon.

Please check out this video: th-cam.com/video/nHpJaE559r4/w-d-xo.html

@@TechMakerAI after doing all this can it run even without pc?

Yes, you can. It is possible to make Raspberry Pi execute the Python program right after startup.

@ryeetzy2624, Yes, Google Gemini API has a free tier. You can use it for free within a limit.

@@TechMakerAI what is the limit?

It depends on the model you choose, see the rate limits here ai.google.dev/pricing

Hi, @ryeetzy2624: Yes. If you change line 39 to "openaitts = False", it will not use the OpenAI TTS. So you do not need OpenAI service. Alternatively, you can use this newer version: th-cam.com/video/b7lRQkJ2xJA/w-d-xo.html or Meta Llamma-3 model at: th-cam.com/video/X8rpOSKDg6I/w-d-xo.html

Hi, @ryeetzy2624, Please check the updated Github page for detailed instructions after that, github.com/techmakerai/Python-OpenAI-API-Voice-Chatbot/

@@TechMakerAI yeah i think it's expired

Hi, @ryeetzy2624: If that is the case, you can try the Google Gemini model at: th-cam.com/video/b7lRQkJ2xJA/w-d-xo.html or the Meta Llamma-3 model at: th-cam.com/video/X8rpOSKDg6I/w-d-xo.html.

Hi, please reach out to me at GitHub

@@TechMakerAI I have followed you and how do I send you a message?

Hi, please reach out to me via my email on my TH-cam channel's "About" page.

That is a great question. The current version of the code allows you to have a conversation with AI. That will help your students with speaking and listening. You can also ask AI to explain grammar questions. I am not sure how you want to teach English? Are you a teacher? Do you teach English to native speakers or to non-English speaking students?

@@TechMakerAI to teach non English speaking student ( Indonesia students).. please help....any link for the code ( any full tutorial about it)... Highly appreciate for your genericity.. God bless U 🙏

Hi, @firmankrisnadi7248, I think you can certainly use it to teach English. For example, you can ask ChatGPT to create a study plan for someone who does not know anything about the English language. You can also ask it to generate some questions about a paragraph to test the students' reading and understanding. Please check out this article for more info: thinkinginenglish.blog/2023/02/27/211-how-to-use-ai-and-chatgpt-to-study-english.

Thank you 👍

Yes, it should work on Raspberry Pi.

It may not be an easy task to use multiple languages in one input (question). But you can certainly ask Gemini to translate from one language to another, see my new video here: th-cam.com/video/n1l81Vv6qQI/w-d-xo.html. Please subscribe to my channel as I have a new video coming up in which I will cover more about this topic.

@@TechMakerAI Thankyou for your reply. Also I had doubt regarding google account creation, on website it says auto payment is not supported but as soon as made my billing account I got a msg : Auto-Pay activated on Kotak Card xXXXX for Google Cloud from 29/05/2024 for INR 75000.00., why so ? Will it auto pay after the given time period ?