Generate Automated Podcast with AI | Open Source Notebook LM

within just couple of seconds now you can create a podcast by AI with two voices which is guest and host about your desired topic drive from your PDF

files or URLs or websites without you manually doing anything or better to say the opsource version of notebook LM then

let's go before we start make sure you subscribe and hit the Bell icon so you will get notified for the next video thank you all right let's get into podcas defy

well what we're going to do here we're going to use this open-source solution to see how we can create a podcast which means it has a host and a guest so two people talking about

uh a topic that it is DED from a PDF that we upload or a website that we provide or M multiple URLs let's say you have a business you have a company you

have products you have uh reports that can be from variety of different topics and you want to create a podcast around all of them that two people are talking about that and you can use that podcast

for different purposes for personal branding for generating leads for sales whatever you want to so as you can see podcas ify says it is an alternative

version to notebook LM so it's an open- Source alternative version then what is notebook LM so before I talk about podcas ify and can use it notebook LM

that recently released by Google which technically is your personal AI research assistant so here you can create a new notebook and then after clicking on new

you need to upload any sources of data you want to chat with you want to do rag here and you can start deep diving in if you can upload or you can get them from

a link if their website YouTube Google Docs Google Slides so on and so forth and then when you upload them you can of start chatting with them asking questions or brain strong with those

documents that you have it's sort of giving you rag option here but just a single use of uploading or connecting here then afterwards after actually

having your data uploaded let me actually open the one that I already uh did before or let me see I think this is

the one I Ed there you go so I had a PDF about MG YouTube channel report this is the one that I got it from a research llm research API that I created a video

out of it previously and then this report is three pages generated by llm about what is my channel what I talk about so I uploaded this it's almost three or four pages of PDF and now I can

start asking questions and stuff plus I can generate a podcast out of it which just a click I already did that and let me show you how it looks

like it's actually pretty realistic take a look okay so let's say you're like me you want to really get machine learning and cloud computing but all the technical stuff I mean who wants to Wade

through all that right it can be a lot it can be so that's what we're doing today it's a deep dive deep dive love it into this report I found on the mg

YouTube channel it's mg Cafe on YouTube if you want to look it up later okay it's almost around 4 minutes but as you can see it's pretty realistic you can see the people are saying uh like a stop

words so now how we can have it open source and give it a try for running it maybe even locally by even local language models or open a models like

any any sort of language model and even any sort of other voice models that they can generate text to voice so podcas ify I found actually just one of many

examples that is available so you can take a look but giving back to podcast it's the maybe most simplest one you can give it a try so how you use it if you strugle scroll all the way

down you just need to to keep install F FM this is actually for audio processing and it's a prerequisites make sure you have python 3.1 or higher and then just

P install podcas ify that's it then you can generate podcast with providing URL or list of URLs these are sources of data that you want to generate podcast

about them and these URLs can be your local PDF path and that's it so let me show you how I run it locally on my Google uh sorry vs code so here I am I

have already installed podcast defy so I just need to import its client as generate podcast and then here this function I'm just using this to create

the display of audio in uh Jupiter notebook here so nothing fancy let me scroll the way down and here the same PDF that I uploaded in notebook LM which

is an report about my YouTube channel I added that as a path inside this URL to generate podcasts about it now by default this uses the open AI voice

capability for generating audio and of course you need to add your open key as environment value here so let me take a look how it looks like welcome to podcast by your personal generative AI

podcast hey everyone and welcome to the show today we're diving deep into the fascinating world of machine imagine a perfect covering ah The crucial deployment phase

so many brilliant I would say it's less realistic compared to notebook LM but what podcas ify has added here is also

you can connect to 11 labs to use the more advanced text to voice capabilities to generate the podcast obviously you need to have the API key added to

environment variable so just quickly show you what is even labs they do have right different ml machine learning servic this is mainly texture speech as you can see I can select the narrator

there are different AI models cloned by voice actors and whether by type here as a sample can be generated so let me show an example the 11 Labs voice generator can

deliver highquality humanlike speech in 32 languages perfect for audiobooks video voiceovers commer yeah I would say the quality is much better so getting

back to to the code now with connecting to event Labs I can report same podcast but over the same PDF but this time with different voices and now let's check out

the result let's welcome to podcast FY your personal machine learning and cloud computing are such Hot Topics right now so what kind of content does the mg

Channel offer well they cover a really impressive range think everything yeah as as you hear I'm covering impressive ranges I'm kidding okay so that was

pretty impressive now again you can plug in with different language models and there are other GitHub repositories I was actually checking one of them which was using llama the 8 billion parameter

and sorry uh I think two and then 80 and even 405 uh billion parameter versions you can have them locally install on your machine to use them to generate these

podcast and using some hugging face open source models for text to voice I was first initially starting to use those but I found that my local machine laptop here is not capable of actually handling

all the you through out of memory issue that's why I use podcas ify because it's calling external language models and even text to voice models but again you can have it all locally on your machine and if you're interested let me know

I'll try to find a way to have that also created in a seate tutorial all right that was quickly about this video I hope you I hope you you find it helpful if yes I would be very thankful if you click on like icon and make sure you

write down your thoughts questions on comment section and the source of all these codes and stuff is added to the Discord Channel Channel which is embedded in the video description below

thank you all [Music]

[Music]