https://www.youtube.com/watch?v=hCEdm9LGBb0
In this tutorial, I’ll show you how to build this automation in two ways:
In my testing, I’ve found that it is indeed possible to build this without writing any code.
So I’ll begin by showing you to do it this way – except we are going to add a single code step, as doing so will eliminate a lot of unnecessary extra costs you’d incur by going completely no-code.
Don’t worry – you’ll be able to completely copy and paste that one code step without needed to understand it.
Click here to jump straight to the no-code portion of the tutorial.
I’m also going to show you a second method, which uses code steps for nearly every part of the automation.
This “code-heavy” method is much more robust, and is the method I’m personally using.
The reason I’m using it is that the no-code method doesn’t currently have a good way of dealing with all the limits of the tools we’ll be working with.
In particular, it’s not good at dealing with ChatGPT’s token limit.
I’ll explain this in more detail later on in the article, but the gist is that ChatGPT can only handle roughly 3,000 words at a time. This limit includes the prompt, transcript (the “context”), and the response.
This means that it can’t natively handle a very long transcript. Want to transcribe and summarize a 1-hour podcast episode? The no-code method can’t handle it.
But the code-heavy method can.
So, if you’re not scared of copying-and-pasting some code blocks, I’d recommend using the code-heavy method. But the choice is up to you, and you can always start with the less-intimidating no-code method just to get your feet wet.
Click here to jump straight to the code-heavy portion of the tutorial.
One note: I’m not going to spend a ton of time deeply explaining the code in this tutorial. However, I do have an extremely detailed tutorial on the Notion API, which does deeply explain the code used: