In less than 50 lines of Python code, I'm going to show you how you can generate NotebookLM-style podcast programs programmatically. The main catch here is that you can use an API to generate the same. See, everybody loves Google's NotebookLM. It's this interface where

you can go upload your PDF, a resume, or any kind of PDF, a paper, or something, and then it would generate a very interesting discussion between two hosts. And it will generate it and give it to you. People are using it for homework, people are using it to listen to

something that they don't understand during their morning walk. So, this is a successful product. The problem is you cannot generate something programmatically. There is no API at this point from Google for you to generate this programmatically. Imagine, like you are a college student or university

student, you want to come in on the weekend and upload all your PDFs and then just generate all these things sequentially. This batch generation is not possible with Google at this point. So, I'm going to teach you how you can do this using a service called Play.ht, and I'm going to

show you the entire code. Like, this code is basically, I think like 50 lines of Python code, I guess, probably. I'm going to share the Google Colab notebook with which I've built this so you can also do the same thing exactly. First of all, you

need to generate your API keys. So, I've already generated my API keys from Play.ht, and then I've added it to my Google Colab secrets. That's the only thing that you have to do other than adding your own PDFs. So, make sure you get the uh

key and the user ID and then add it to your Google Colab Secrets. If you're running this on a local computer, then you can use something like import os. environ and then you can just give whatever that name that you want and you can add the key that you want. So

if you're running this on local, you can do this thing. But if you're running this on Google Colab, you can just exactly do whatever I'm doing right now. After this is done, I've linked the documentation for you to go play with this, but if you're

using this notebook, you don't need documentation. So, we need two libraries: primarily 'requests' and 'os'. Um, this is if you're trying to read the keys locally, but because I'm on Google Colab, I'm going to use this. Next, you have to define the PDF from which you want to generate a

podcast. The demo I'm going to show you is from the Amazon shareholder letter that Jeff Bezos wrote in 2007, but the demo that is what I'm going to use to generate. But I've already created something, and which is I think based on

um, the 7even Powers by Hamilton Helmer, like it's a book. And then I'm going to play the audio for that now. After you have defined what is the source file, next you have to go and then hit this API. The API that you have to hit is uh this: api.

play. /i/ V1 play notes. And now you set the keys and after that, you are going to basically set up the header for the API request. Once that is done, the main juice is here. So, you have to specify the source file URL, which is what the PDF

Here is, and then you have to say what kind of, uh, synthesis style you want. We are doing a podcast because that is a popular format, but this API endpoint supports things like children's stories, executive summaries, and a bunch of other things. You can literally go to the endpoint here

and then check what other, um, types of audio that you can generate. Some are like single host, some are multiple host. In our case, this is a two-host system, that means we need two voices. We're going to take two voices, Voice one, uh, we're going to call The Voice Angelo, and

then Voice two, we're going to call The Voice DD. You can give any name that you want. For example, if you want to, uh, create like a podcast for your company, you can like imagine you have got like the CMO and CTO, so you can have the CMO

name and the CTO name, so it's up to what you want to do. And I think once you have the voices ready, then all you have to do is you have to send a post request with the given headers and files, and then you have to check if the response is 201. If

the response is 201, then your request is successful. If not, then you're going to get an error message with whatever the error message is. But once the response is successful, you're going to get an ID. So if I go particularly run this, I hope I don't get an error because okay, I did

not, you run this, that's my bad. So I'm going to run this first and then I'm going to run this, and the moment I run this, you can see that the request has been sent successfully and there is a play note ID generated for whatever that

I wanted to create it. Might take a few minutes, so instead of keeping you waiting, what I've done is I've already created a play node which is stored inside this play node, and I'm going to play that for you. So the way I'm going

to play that is you have the play note ID. So let's run this after you have the play note ID. So you have to go hit this endpoint with the play note ID and that will ultimately give you the status. For example, if I were to go use it, let me

stop this for a moment. Let's say this is what I am uh generating, okay? So if this is my play note ID, play note ID, then if I run this it would most likely say that this is currently, yeah, you please wait while your play note is being generated.

Try again later because this is still going on. It takes a couple of minutes for it to generate, but in my case I've already generated this, so that if I use this as a play note ID and then I hit this end point to pull if it has been

successful or not, if it is successful, I also get the final audio clip. So in this case, I've got the audio clip. So then all I have to do is download the audio clip which I'm using like the Linux command 'wget' and I'm storing this here in my

local computer or within my Google Colab environment. And after I do this thing, all I'm going to do is I'm going to use the Jupyter notebook or IPython widget just to play the audio. So I can go here and then play the audio. I'm

going to play the audio for you to listen hey everyone welcome back to Power Play navigating business strategy I'm Dei your co-host for today and I'm Angela so glad you're all tuning in Dei I've been itching to talk about today's topic oh me too we're diving

into Hamilton Helmer's seven Powers the foundations of business strategy uh-huh that book really shook up my understanding of competitive advantage and customer relationships that's a fair point but Helmer doesn't ignore those elements he incorporates them with as you can see here we have got two

very successful hosts uh Dei and Angela discussing the book that we have got here which is um which is the seven powers from Hamilton Helmer and they're like literally discussing the book which is exactly what you want whether it is your homework whether it is a resume

whether it is an intern document whether it is like a business executive summary whatever you want you can generate it like a podcast just like Notebook LM using this endpoint uh I guess they give a particular set of free credits for you to get started so you can go here and

get started with this but the main idea here is that can you generate Notebook LM style podcast programmatically imagine you have got 100 PDFs you want to send it as a batch and then generate it yes it is possible you don't have to

deal with the headache of LLMs you don't have to uh deal with how do you take this text of PDF and then convert into text you don't have to deal with any of this all you have to do is have one PDF link give that PDF link URL anywhere

that you find on the internet, give that to this endpoint in this particular parameter, source file URL, that actually goes inside your um this this request parameter and all you can do is generate the ultimate final podcast style or notebook LM style podcast in an MP3 so you can run your own radio

station within your University or within your company and uh the potential is quite enormous. I'm I'm pretty sure a lot of the internet is going to be flooded with this content, but for now if you're a Python programmer and you wanted a way to generate notebook LM style podcast

using an API, this is the code for you. Link to the code in the YouTube description. Please let let me know what you feel about this kind of solution. See you in another video. Happy prompt day.

NotebookLM is PERFECT, but... 💥Create AI Podcasts with NotebookLM API Tutorial 💥

NotebookLM is PERFECT, but... 💥Create AI Podcasts with NotebookLM API Tutorial 💥