Create Structure Out Of Messy Texts

So, I would like to write an article. About. Using. I'm chat, GPT API inside of google sheets Made a simple guide on how to Texas, the GPT API inside of Google Sheets in that article. I've also described some simple use cases Like doing public descriptions. And sentiment analysis. Like you said, GPT API is a really good at It's really good at Doing that spurts. Without the API. It's hard to do a lot of them in your time. Now, that the price Request as significantly decreased it's way way more interesting. To do a lot of work at once is sent out a lot of bumps. And so with, The very simple script you can do most of that, in very little With very little effort. So basically type in the function, Reference themselves. And drag. The function down through a bunch of extra cells. Evans in no time, you have It's on the same prompt on a lot of different cells. In an affordable way too. 10 times cheaper is significant. Not as we go first. Why? Or once we wrote about and a previous article and Well, it's pretty cool this weekend. Enable this We cannot move home to. What the actual article could be about, which is Nothing. Automated article generation based on the Prescription. It could be made by anyone. So, basically, you start with, either on transcription or an audio file, I'm talking for 10-15 minutes. Ferrets and heaven, instant. Fully written article on. I want to explain step by step, how to do this. So far as Determined stats. It takes about. For distinct. Steps. The first step. Us to do with. The recording an automated transcription. But I know there's two. Easy ways to do this. There's more ways. Probably some easy on stew, but they're still examples that are quite clear that work. Well, Let's give a decent results in very little time. Without requiring anything complicates it's like programming. The first is the Google recorder app. This app allows you to record and have what you say, automatically transcribed. It works pretty well. It's not perfect. And the second one is open a eyes. Whisper. With the whisper API, you can just upload an audio file and for A very cheap price, it'll transcribe it Really, really well. There's also, The whisper normally require some programming, but There's a little. It for you. Are you going out there wants to do this? But doesn't really know how to go. Go to. Open a ice playgrounds. On the playground. There's a page where you can. You can do prompts using. 253. And any of the other models, That's the option. Lick on the microphone at the top, right? This will actually allow you to use whisper within a browser. Probably take some of your tokens but it's stupid cheap anyway, so I wouldn't really. Go ahead. Start talking. Conference, you're done stop. And you'll notice it's a perfectly transcribed, text. For you to use to copy and paste. So, All my friends who don't know how to program. We don't really know how to use my phone. And don't want a lot of set of time. There you go. So that step one. Recording. Just driving. At this point. Go to transcription. You want to do now? Here comes step two, clean up the trans person. This part is, The problems you need to use. Something along the lines of you are transcriber. You need to clean up the following. Followed by the transcription. You also need to remove some repetition, but you have to be a bit careful there to clearly instruct If you want to retain as much information as possible, because ABI really likes to. Output. Small amount of specs which would then remove all the context. Why don't you have a simple prompt? That remove some of the repetition, it cleans up the ass and arms in the text. But whenever you say, Or. Or make mistakes. Another crucial part in this step is to add one or two line of context. For example, I'm talking about jet GPT now. It doesn't always work. Perfectly. So be scoopy changed for bees. Prescriptions can make simple semantic, mistakes and Westboro works better than Google record if it's still. You have to be a bit careful if that's Into the next step. I meant to say, whisper, Makes less mistakes. Google recorder. Whisper makes less mistake than google recorder. I was going to say, Is transcribing software. Can make little semantic mistakes and feeding it through a problem for cleaning uploads. So I think we've done the first couple steps first, two steps specific. Now, it's time. Beautiful too. Now, it's time to move on to step number three. And step number three is, Determining the structure that's what we've used. So, I want this to be able to, Brighter newsletter. An article writer step by step guide. Provides technical documentation. Are something for within a company. Like documentation for a company. Local block post. Yeah, blocking would be a nice example. Login, with a G. Blogging. Yes. How would you go about selecting it? I guess you could tell. Ask for it yourself, but with The power of. GPT. My daughter option is to simply ask. What sort of? What's with the transcription fit? So if the transcription is about something technical, it could Through a prompts suggest. Publishers, or Magazines are a type of writing. So, Right now, I'm talking about using jet GPT API. In Google sheets, which is a fairly technical subject and They might be some bloggers or they might be some newsletter, substick newsletters. That's what fit this. Or something, like medium or Wires. Or let's go. Or an author like Dan Shepherd, from every Quit, write something about this. And then you could ask it if you've got something like that writing. About. Or if you've got an article from one of those, what would they generally structure? The article is and suddenly you've got an outline for an article. That fits your The topic you're talking about. Once you have to structure, you're looking for, You can then use a prompt. To describe. Every section and the function of it section. This is absolutely crucial, because This is going to be the basis for the article that you're going to write. With a simple split function, you can then split up. The difference. Sections of a different cells. With the steps, you end up with. An outline of an article with different sections, then you can use a prompt to send those sections. Why not Santa sections? I mean Then you can use a prompt. That adds. A brief summary of the content. That's will be contained in those sections based on the I'm cleaned up. Transcription. This is a crucial step. Because this will be your guiding Because of version of, The article that you're going to write this is going to be the outline, that's Will be used as a mold to fit the transcription in. South Houston out a little bit. This means you're going to end up with The outline of an article with sections and titles with the function of section. And then Short description of what the content is going to be based on the transcription you've made. With a split function, you can then cut those up and reference those from writing your article. The model really likes to output small amounts of text and with small amine around, 400 words. Which is not nearly enough. Water, trying to achieve her. Which means that's with a split function in Google sheet. You can cut it up into smaller parts and now we get to the next step which is writing the actual article. So for riding the article, you're going to need three things which is the outline of the article. The transcription and then, I'm going to watch the section that it needs to write. It needs to know what the next one is going to be. Otherwise, it's going to revise parts again and again. Which is not over looking for here. In the end, you're going to end up with parts of an article with some overlap that you can put together. You might need to remove some parts that are written double, but if you prompt it, well then they should be very easy. I should can. Includes in the output of the prompt. As you can put in the section title. In the output of the broncho, you know, exactly when it starts to overlap and it stops overlapping. Yes. So this is the step by step guide on how to achieve all this. And also achieve a full article based on, A transcription and that's it for now.