- Aron Levin
- Posts
- How to "write like you talk", turn video into text and outsmart 95% of the population
How to "write like you talk", turn video into text and outsmart 95% of the population
A simple step-by-step process that'll teach you how to write like you talk and turn video or audio into an engaging article (or any form of written copy, really) in just minutes.
If you simply manage to write in spoken language, you'll be ahead of 95% of writers. And it's so easy to do: just don't let a sentence through unless it's the way you'd say it to a friend – Paul Graham, Write Like You Talk
Remember the GaryVee Content Model? Fan or not… the basic premise of the model is to take a long-form piece of content and break it down into smaller pieces of content like clips, quotes, takeaways, and summaries that you can distribute across different platforms, like social media, podcasts, blogs, and other channels.
But as the saying goes, “In theory, theory and practice are the same. In practice, they are not.” And that’s the thing. It’s great in theory‚ but not very practical.
Unless you’re a six-figure keynote speaker with a 30+ person team of content producers at your disposal, it’s very challenging and time-consuming to turn a 60-minute video into 30 pieces of unique content overnight.
Until you have the right tools at your disposal, that is. In this case nothing but free software, a text editor, and ChatGPT.
As for today, I’ll walk you through a simple step-by-step process that’ll show you how to write like you talk and turn video or audio into an engaging article (or any form of written copy, really) in just minutes.
Showtime
There’s some tech involved here, but there’s no need to be alarmed. It’s simple stuff.
First, find a video or (audio file) featuring you, yes you, speaking on a topic of your choice. Alternatively, jot down a few talking points, and open your favorite audio or video recording app (like Quicktime on your Mac or the voice recorder on your phone). No need to get fancy.
I don’t know about you… but I’ve found walking or driving to be when it’s both natural and easy for me to get into a state of flow and record my ideas.
But… as of this very moment, there’s heavy rain, and my car is double-parked, so we’ll have to do with this video from some 6 years ago featuring myself speaking on stage at an event. The video is 27 minutes long. Perfect length and context for our project.
There are several online tools (and free apps) that’ll transcribe your video or audio recording to a text file. We’ll use a free one that runs in the terminal on your Mac called youtube-dl. This will allow you to download subtitles from any YouTube video (and most other online video services), with the added benefit of never running any limitations, paywalls, or pesky popups… ever. How cool is that? Plus, it’ll make you look really smart.
How to install youtube-dl
First, you need to install Homebrew. Homebrew is a package manager for Mac that makes it easy to install and manage command-line tools like YouTube-dl. To install Homebrew, open the Terminal and paste the following command:
/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/master/install.sh)"
Once Homebrew is installed, you can use it to install YouTube-dl. To do this, run the following command in the Terminal:
brew install youtube-dl
Once YouTube-dl is installed, you can use it to download the subtitles from a YouTube video and most other video services. To do this, you will need the URL of the YouTube video that you want to download the subtitles from.
To download the subtitles from a YouTube video, run the following command in the Terminal, replacing VIDEO_URL with the URL of the YouTube video within ''
youtube-dl --write-auto-sub --skip-download 'VIDEO_URL'
This will download the subtitles as a .vtt file. The file will be saved in the current directory, with the filename in the format uploader.title.id.vtt.
22,144 characters of mess
I noticed that YouTube, for some reason, will include the transcription twice and a lot of other data timestamp that we won’t need. Clean up the file and keep the first part, and then save the file.
The next step is to split the text file into several text files and keep each prompt within the 15,000-character limit. I’ve explained how to do this here.
My transcription is 22,144 characters, and I end up with two separate text files.
Now for the fun part.
A content producer at my disposal
Head over to chatGPT and follow the steps below.
I’d encourage you to play around with the formatting, but I’ve found, through trial and error, that context and detailed instructions improve accuracy. Just like instructing a team member or assistant to help with a task, you’ll want to leave as little room as possible for error. It’s better to overcommunicate than to assume that context is understood. Generally speaking, of course.
You can copy this exact prompt if you’d like and replace the bold part.
Prompt: Now follow these exact instructions. Let's think step by step: I'm going to need your help analyzing a text. The text is a transcript from [a YouTube video featuring me (Aron Levin) speaking for marketers at Adobe's conference.]Because it's a subtitle file, the text doesn't have punctuation, paragraphs or proper formatting. Because of its length, I will be sharing the transcript across two prompts. I will label each prompt "Part [1-2]:". When you have received the prompt, just type out "Got it, please share the next part.". When I have shared all parts, I want you to confirm this by typing out "I received all parts and am now ready to analyze the text.". Do not analyze the text until I give you further instructions. Ok?
Answer: Sure, I can help you with that. I will wait for you to share the text with me in two separate prompts, labeled "Part 1" and "Part 2". When I receive each prompt, I will respond with "Got it, please share the next part.". When you have shared both parts, I will confirm that I have received both parts by responding with "I received both parts and am now ready to analyze the text.". I will not begin analyzing the text until you give me further instructions. Is that what you had in mind?
I proceed with “Yes” and share my two-part transcription. I start each prompt with “Part 1” and “Part 2”:
Answer: I received both parts and am now ready to analyze the text. Is there anything specific you would like me to help with?
Prompt: Let's start by making sure that the text is fully understood. Do you have any specific questions for me based on your analysis of the text?
Answer: Based on my analysis of the text, it seems to be a transcript of a presentation given by someone named Aron Levin at a conference for marketers at Adobe. The presentation discusses the shift of advertising spend from traditional media companies to a long tail of creative individuals, and how brands can take advantage of this trend by engaging with the creative community and producing relatable and authentic content. The presentation also includes references to a YouTube video and a campaign by Uber.Is this a correct understanding of the content of the text?
Spot on!
Brainstorming ideas
What follows are a few prompts exploring content ideas. You don’t have to do this if you already have an idea in mind. Here’s what I came up with:
A summary or overview of the main points and takeaways from the presentation
An analysis or critique of the ideas presented in the presentation
A list of notable quotes from the speech
A list of actionable tips or strategies for brands looking to engage with the creative community and produce authentic content
An article or blog post exploring the shift towards a long tail of creative individuals in the advertising industry and its implications for brands
A case study or success story highlighting a brand that has effectively engaged with the creative community and produced authentic content
Oh, look at that. “Analysis or critique of your content,” or even detailed feedback on a presentation. That’s an interesting AI use case. Public speaking is, after all, both a multi-billion dollar industry and commonly cited as the most common phobia in the world. And there’s an entire AI-powered coaching aspect too...
… But ADD brain be damned. Step aside! We've got work to do, so let’s get back to business.
Prompt: Let's write an article based on the transcript. Can you give me a few suggested titles?
Sure! Here are a few suggested titles for an article based on the transcript:"The Rise of the Long Tail: How Creative Individuals are Disrupting the Advertising Industry""Authenticity is Key: Insights from Aron Levin on Engaging with the Creative Community""The Future of Advertising: How Brands Can Thrive in a World of Influencers""From Super Bowl to Social Media: How Uber Revolutionized its Marketing Strategy""The Power of Relatability: Lessons from Aron Levin on Connecting with Your Audience"These are just a few ideas, and there may be other titles that could be effective for the article. Let me know if any of these suggestions resonate with you, or if you have any other ideas for the title of the article.
We have a long way to go, in my opinion, before copywriters are replaced by machines, but this ain’t bad. “From Super Bowl to Social Media” is a decent and catchy phrase. I like the first suggestion.
Prompt: Let's go with the first suggestion. Let's make it a thought-provoking thought-leader piece. Can you list a few bullet points on what the article could cover?
To spare you the back and forth, I’ll highlight a couple of key prompts and suggestions that followed: Be specific about the article length and structure. Ask to incorporate the tone of voice, vocabulary, etc from the video so that it follows your unique tone of voice, style, pattern, flow, and personality. etc.
The result? About 5 minutes later I end up with a 700-word draft (Google Doc for reference.) accurately titled The Rise of the Long Tail: How Creative Individuals are Disrupting the Advertising Industry.
A cornerstone for short-form content
From here, you can repeat the steps we’ve taken and create more drafts on different topics and angles, and let the draft function as a cornerstone for other content such as LinkedIn posts, emails, etcetera.
Example prompt: Generate 5 thought provoking and polarizing tweets from the article. Make them statements, not questions.
"The future of advertising belongs to creative individuals with niche audiences, not traditional media powerhouses #longtail #disruption"
"Brands need to stop trying to reach everyone and focus on connecting with niche audiences for real engagement #nichetargeting #authenticity"
"The rise of social media has given individuals more publishing power than ever before, and it's changing the game for brands #influencermarketing"
"Authenticity and relatability are more important than reach when it comes to advertising and building brand trust #authenticcontent"
"The advertising industry is being disrupted by creative individuals with niche influence, and brands need to adapt or be left behind #disruption #innovation"
You get the idea.
Flavor and personality
It's inevitable, in my opinion, that 99.99% of all written content online will be AI-generated. Suddenly, everyone’s a writer, and you won’t be the only one churning out AI-assisted or even entirely AI-generated content.
But … there's another side to this, and where things get interesting: AI-generated text doesn't have a distinct tone of voice or character. It’s, by default, generic, boring, and bland. And since most companies and brands already sound alike… it’ll only get worse.
Meaning that a unique, authentic, and recognizable tone of voice and style will be more important than ever before. See, when the robots do *take over* the internet, we’ll gravitate to people with an actual personality, no matter how obstreperous* or voluble* they are (** two new words that I learned today on Quora, in reference to that very same six-figure keynote speaker.).
On that note, your draft will likely need more personality.
My writing is more relaxed, conversational, and informal. I mix technical and industry-specific terms with colloquial and everyday language (I also tend to use parenthesis to provide commentary on observation). From time to time, there’s also the occasional use of self-deprecating humor, wit, sarcasm, and a dash of profanity.
How do I know? I used OpenAI (of course) to analyze a couple of old articles and describe my style of writing. I even took a stab at incorporating those exact traits in a rewrite of the draft, but it actually made it… worse. There was, indeed, lots of flavor and personality, but something was off. Very tricky to get the right balance and mimic my exact style of writing (and nope, we’re not going down the fine-tuning rabbit hole just yet.)
And it’s not the only challenge. Whether in written or spoken form, our stories are told from past experiences, memories and observations in culture. We offer opinions and perspectives without policies or guardrails. And to resonate with your audience, in order to matter, you need to communicate what you stand for.
Keep this in mind when you’re at the initial stage of recording your audio or video content; I’ve found, at least for now, that it’s incredibly challenging to fix with the assistance of AI at a later stage of your creative process.
– Aron