Sound human,
keep the listener.
Here is a quiet machine that makes my podcast episodes sound like a real person, not an AI. My scripts are written by an AI and turned into audio by a clone of my voice, then posted to Spotify and Apple Podcasts. But before any of that happens, this machine goes into my script repository and humanises every one. First it rewrites the text for reading out loud, the way a person actually speaks, not the way an article is written. Then it strips out the tell-tale phrases that scream AI, the ones we can all spot in a second. The result is stupendous. It sounds like someone speaking normally, so listeners stay longer, come back, and click the links in every episode that lead to my paid work.
People do not write the way they speak. AI writes an article, then reads it like one.
AI is good at writing, but it writes a podcast script the way it would write a blog post. Neat sentences, tidy phrasing, none of the little turns a real person uses out loud. Read that on a page and it is fine. Say it into a microphone and it sounds off, and listeners feel it even if they cannot name it. So I built a machine that fixes this before a single word is recorded. It rewrites every script for the ear, then strips out the phrases that scream AI. Here is who it is for, what goes wrong without it, how it works, and what you get back.
Who it’s for
Anyone using AI to write audio content, podcasters, course creators, marketers who post talks and clips. If you write a script with AI and then speak it, whether in your own voice or a cloned one, this is for you. It matters most if you post often, because the AI tells pile up fast and listeners start to feel that something is not quite human.
What goes wrong
Without it, your AI script reads like an article being read aloud. Stiff phrasing, the same tired openings, the words everyone now knows come from a machine. Listeners drop off, not because the ideas are bad, but because it does not sound like a person, and people do not stay with something that feels fake.
How the machine works
It goes into your script repository and takes each one. First it rewrites the text for reading out loud, swapping article phrasing for spoken language. Then it strips out the tell-tale phrases that scream AI. You write nothing by hand. The machine cleans every script and hands the clean version to your voice.
What you get back
Episodes that sound like a real person talking, so nobody suspects they are automated. Listeners stay longer, come back, and click the links you place in each one. A more human episode keeps a few more listeners to the end and earns a few more clicks, and across a growing back-catalogue, that compounds.
I did not want a robot reading my podcast. I wanted it to sound like me, talking.
This automation greatly improves the quality of my podcast episodes and makes them more relatable, easier and more pleasant to listen to. That grows my audience and my listen time. Here is the setup behind it. I have a repository where all my podcast scripts live. An AI writes them, and a clone of my voice turns them into audio that gets posted to Spotify and Apple Podcasts. It all runs without me.
But before that happens, I want to make sure the podcasts sound as human as possible, and that everything that screams AI is removed. The problem is simple. AI tends to write a script as if it were a blog or a website article. People do not write the way they speak. So a script that reads perfectly on a page sounds stiff and lifeless the moment a voice reads it out loud.
So the machine goes through every script and runs it into another automation whose only job is to humanise the text as much as possible. First, it optimises the script for reading out loud, as if someone were reading from a teleprompter, removing written-article expressions and replacing them with spoken language. Then it strips out the tell-tale expressions that scream AI, because AI keeps writing things the same way and we can all spot it.
The result is stupendous. It sounds very human, and honestly nobody knows these are automated podcasts using my cloned voice, because it sounds like someone just speaking normally. The bot does all of this on its own, going into my repository and optimising every script sitting there, so I never touch a word of it myself.
The numbers here are kept deliberately modest. A more human episode simply keeps a few more listeners to the end and earns a few more clicks on the links inside it than a stiff one would. That edge is small on one episode, but I publish many, and the back-catalogue keeps being discovered for years. A small edge on every episode, compounding across a growing library, quietly turns into real revenue.
Proof point: I have documented how I run my whole business on autopilot and the income streams behind it on YouTube, in a day in my life and in my 28 income streams breakdown, so the conservative numbers on this page are checkable.
Two passes that turn an AI article into a human-sounding episode
What made this work was accepting one simple truth. People do not write the way they speak, and AI writes for the page, not the ear. So the fix is two clear passes over every script before it ever reaches a voice. The first pass rewrites the words for reading out loud, the way a person actually talks. The second pass hunts down the phrases that give AI away and removes them. Once both passes run, a stiff article becomes something that sounds like a real person, and that is the whole difference between a listener staying and a listener leaving.
Write for the ear, not the page
The first pass rewrites the script the way a person actually speaks. AI writes in full, tidy sentences, the kind you read in an article, with phrasing nobody uses out loud. This pass swaps all of that for spoken language. It shortens sentences, adds the little turns a real person makes when they talk, and reshapes the script as if it were meant to be read off a teleprompter. The words stop being something you read and start being something you say. When a voice reads the result, it flows naturally, with the rhythm of real speech, and the listener relaxes because it sounds like a human sitting across from them, not a document being narrated.
Strip out the tell-tale AI phrases
The second pass hunts the tells. AI keeps writing the same way, reaching for the same openings, the same tidy transitions, the same handful of words we have all now learned to spot. The moment one of those lands in an episode, a listener feels it, even if they cannot name why. This pass finds every one of them and removes it. It clears the stock phrases, the repeated patterns, the giveaway vocabulary, and replaces them with something plain and human. What is left has none of the fingerprints that make a listener think a machine wrote this. It just sounds like a person who knows their subject, talking about it in their own words.
Hand the clean script to your voice
Once both passes are done, the clean script goes straight to the voice. In my case that is a clone of my own voice, but it works the same with a real one. Because the words were already shaped for speech and cleared of AI tells, the recording sounds effortless. The pauses fall in the right places, the phrasing carries the natural rhythm of talking, and nothing trips the ear. Nobody listening suspects a machine wrote the script or that a cloned voice read it. It simply sounds like a normal episode. That is what keeps people listening to the end, and coming back for the next one.
Once both passes are in place, the guessing stops. Every script that lands in the repository gets rewritten for the ear, cleared of its AI tells, and handed clean to the voice. You do nothing. The machine runs each one the same careful way, so every episode sounds human, holds the listener, and earns the clicks inside it. A more human episode adds a little on every publish, and across a growing library, that quietly compounds.
Before the system
- AI scripts that read like articles being narrated out loud
- The same tired openings and phrases in every episode
- Listeners dropping off because it did not sound human
- Editing scripts by hand to make them speakable, one by one
- Fewer clicks on my links because people left early
After the system
- Every script rewritten for the ear, the way a person speaks
- The tell-tale AI phrases found and stripped out for good
- Episodes that sound like a real person, nobody suspects a thing
- The whole repository humanised on its own, hands off
- More listen time and more clicks to my paid work
Prompt 1: rewrite an AI script for the ear
The first job is turning writing into speech. AI writes for the page, so the script has to be reshaped for how a person actually talks. Use this prompt to rewrite any script so it reads naturally out loud, like a teleprompter, not an article.
Read-it-aloud rewriter
Act as a spoken-word editor for podcasts. I have a script written by an AI that reads like a blog article, and I want it rewritten so it sounds natural when read out loud in my own voice. About my show: [describe your podcast, your topic, and the way you normally talk]. Rewrite the script for the ear: shorten sentences, swap written-article phrasing for how a person actually speaks, add the natural turns and rhythm of real talking, and keep every idea intact. Give me the rewritten script, then one line on the main changes you made and why they help it sound human.
The output is a script shaped for speech, not the page. Get this right and the words stop sounding like a document being narrated and start sounding like you, talking to one person.
Prompt 2: find and strip the tell-tale AI phrases
Even a well-written script carries fingerprints. AI reaches for the same openings and words, and listeners have learned to spot them. Use this prompt to hunt down every tell in a script and replace it with something plain and human.
AI-tell remover
Act as an editor who removes the signs that a text was written by AI. I have a podcast script and I want every phrase that gives away a machine wrote it found and replaced with plain, human wording. About my content: [describe the tone and the audience so the fixes match your voice]. Find the tells: the stock openings, the tidy transitions, the repeated patterns, and the vocabulary people now recognise as AI. Replace each one with how a real person would say it, without losing the meaning. Give me the cleaned script, then a short list of the tells you removed.
The output is a script with the machine fingerprints gone. Run this once and nothing in the episode makes a listener think a bot wrote it, because the giveaways are all cleared out.
The 3-minute overview of how this works
Before the build steps, watch this short overview. It’s the exact video from the Automations Made Easy page, and it walks through the mechanics behind machines like this one. 1,000+ students have used these mechanics to save two hours a day, with zero coding.
Want to learn the mechanics behind content automations like this?
I teach the same mechanics that power this podcast humaniser inside Automations Made Easy. 1,000+ students have used these mechanics to save two hours a day and build little machines that quietly handle the boring work. No coding required.
Week one: your first scripts cleaned, your first human episode.
Most people put off building something like this because it sounds technical, so they keep hand-editing scripts forever or publishing stiff ones. That is exactly why this machine matters. A humaniser is meant to start simple and run quietly, not arrive perfect. In week one you point it at a handful of scripts, let it rewrite them for the ear and strip the AI tells, then listen to the cloned voice read one, so you hear the difference before you scale it.
Week one looks like this. On Monday you take one AI-written script and run it through the first pass, the rewrite for reading out loud. Then you run the second pass that strips the tell-tale AI phrases. By midweek you set the rules so both passes apply the same way to every script. By the end of the week you have your first fully humanised episode recorded in the cloned voice, with the machine already working through the rest of the repository on its own.
The point of the first week is not how many scripts you clean. The point is to confirm the machine rewrites for the ear, removes the AI tells, and hands a clean script to your voice. Once that is locked, every new script that lands in the repository gets rewritten, cleared, and voiced on its own, and your episodes keep sounding human while you do nothing at all.
From there the maths is simple and conservative. A more human episode keeps a few more listeners to the end and earns a few more clicks on the links inside it than a stiff one would. That edge is small on one episode, but you publish many, and the back-catalogue keeps being discovered for years. Over three to five years, those small edges stack across a growing library and quietly compound into real revenue, all from episodes that clean themselves.
All of this runs while you work, sleep, or record your next episode. The machine opens the repository, rewrites each script for the ear, strips the AI tells, and hands the clean version to your voice. No more hand-editing scripts, no more stiff episodes, no more listeners leaving because it sounded fake. Your whole library sounds human on its own, and nobody suspects a machine had anything to do with it.
Prompt 3: build the rules that humanise every script the same way
Doing this once is easy. Doing it the same way across a whole repository is the real win. Use this prompt to write a clear set of rules the machine follows on every script, so every episode comes out consistent.
Humanising rulebook
Act as a content systems advisor. I want to humanise every AI-written script in my repository automatically, with the same standard applied to each one, not by hand. About my library: [describe how many scripts you have and how often new ones arrive]. Write the rulebook: how to rewrite for the ear, which AI tells to always remove, how to keep my voice consistent across episodes, and how to handle a script that is already fairly natural. For each rule, one line on why it keeps the episodes sounding human.
The output is the standard the machine applies to every script. Settle it once and each episode in the repository gets humanised the same reliable way, whether you are watching or not.
Prompt 4: place the links that turn listeners into buyers
A human-sounding episode only pays off if it points somewhere. The whole reason retention matters is the link inside each episode. Use this prompt to place calls and links naturally in the script, so they feel like part of the talk, not an ad.
Natural link placement writer
Act as a conversion copywriter for audio. I have a clean, human-sounding podcast script and I want to place a link to my paid program and my long-form content inside it, without it feeling like an advert. About my offer: [describe your paid program, your long-form content, and where a listener should go next]. Write the placements: where in the episode a mention fits naturally, how to phrase it so it sounds like me talking rather than reading an ad, and how many times to say it without wearing the listener out. Give me the lines to drop in and where they go.
The output is a set of natural mentions that send listeners to your paid work. Get it right and every episode that keeps someone to the end also gives them a clear reason to click.
The exact build, step by step
Point the machine at your script repository
Start where the scripts already live. You have a repository full of AI-written podcast scripts, so you point the machine at it and let it pick up every one. This single connection sets the whole thing up, because from here the machine handles each script without you opening a file. This is the step most people never automate, so they end up editing scripts one at a time forever. When the machine has the whole repository, every script that lands there, now and in the future, gets cleaned on its own.
Rewrite each script for reading out loud
Here is the first pass. The machine takes an AI script that reads like an article and rewrites it for the ear. It shortens the long, tidy sentences, swaps written-article phrasing for the way a person actually speaks, and reshapes the whole thing as if it were meant to be read off a teleprompter. Nothing about the meaning changes, only how it sounds when spoken. By the time this pass is done, the script no longer reads like a document. It reads like someone talking, which is exactly what a voice needs to sound natural.
Strip out the phrases that scream AI
Now the second pass. AI keeps writing the same way, reaching for the same openings and the same handful of words we can all spot in a second. The machine hunts every one of those tells and removes it. It clears the stock phrases, the repeated patterns, and the giveaway vocabulary, then replaces them with something plain and human. This is the quiet step that makes the difference. A listener may not be able to say why an episode sounds fake, but they feel it, and this pass removes the exact things that trigger that feeling.
Hand the clean script to the cloned voice
Now the clean script goes to the voice. In my case that is a clone of my own voice, but it works the same with any voice. Because the words were already shaped for speech and cleared of AI tells, the recording sounds effortless. The pauses land in the right places, the phrasing carries the rhythm of real talking, and nothing trips the ear. The episode that comes out sounds like a normal person speaking, and honestly nobody knows it is an automated podcast using a cloned voice, because it just sounds like someone talking.
Let the bot work through the whole repository
This is where it stops being a task and becomes a machine. The bot does not clean one script and wait for you. It works through the whole repository on its own, taking each script in turn, running both passes, and handing the clean version to the voice. New scripts that the AI writes tomorrow get picked up and cleaned the same way. You are not in the loop at all. The library humanises itself in the background while you do other things, and every episode that comes out of it holds the same human standard as the last.
Watch the retention and the clicks climb
Final piece, and the one that pays. A more human episode keeps a few more listeners to the end, and every episode has links that send people to my paid program and my long-form content. So the better the podcast sounds, the more people stay, the more people see those links, and the more people click through to what I sell. These are not hunches. A relaxed listener who feels they are hearing a real person is simply more likely to keep listening and act. Every well-made episode adds a small edge, and across a growing back-catalogue that keeps being discovered, those edges stack into real numbers.
Open the repository
The machine goes into your script repository and picks up every AI-written script waiting there.
Rewrite for the ear
It reshapes each script for reading out loud, swapping article phrasing for how a person speaks.
Strip the AI tells
It hunts the phrases that scream AI and replaces them with plain, human wording.
Voice and publish
The cloned voice reads the clean script, and the human-sounding episode goes out to your feeds.
Build this podcast humaniser inside the same playbook 1,000+ students use
Automations Made Easy teaches the mechanics behind content machines like this one. Step by step, no code, plain English. Save two hours a day and own little machines that make your content sound human on its own, so you spend your time creating instead of hand-editing every script.
The six months after I switched it on
Here is the shape of the first six months after I turned the podcast humaniser on. The line tracks the extra value from more human episodes, the small edge each one earns in listen time and link clicks over a stiff one, which is exactly how this machine pays off in practice.
Monthly value from human-sounding episodes
Three things matter on this chart. The value climbs steadily as more human episodes publish and the back-catalogue keeps being found, not in a spike. The gains come from episodes the machine cleaned for free, without you editing a single script. And every one of those months happens while the machine humanises your scripts and hands them to your voice for you.
What other students built
I teach the simple skills behind machines like this in Automations Made Easy. Students who built their own version sent back what changed in their first month.
“My old episodes read like a robot narrating a blog. After the rewrite pass, a listener messaged to say it finally sounded like me. My average listen time went up and stayed up.”
“The AI tells were killing me and I could not even see them. The machine stripped them all out. Now nobody guesses my scripts are written by AI, and my link clicks per episode went up.”
“I run a cloned voice like Martin does. Before this, people said something felt off. After, the comments changed to how did you record all this so fast. The whole library sounds human now.”
“Knowing week one was just cleaning one script and hearing it back kept it easy. By month two my whole repository was humanising itself and I had stopped editing scripts by hand entirely.”
What’s inside Automations Made Easy
AME isn’t a library of pre-built automations. Every business is slightly different. What’s reusable across all of them is the underlying mechanics: how to set up little machines that listen, write, and follow up while you sleep, and how to wire the pieces together without writing code.
The program walks you through six modules: The Right Tools (the cost-effective, no-code stack I actually use), Task Selection Mastery (which automations are worth building first), Design Secrets (mapping an automation before you build it), Zero to Hero (complete beginner to confident automator), Real-World Application (we build a full automation together, end to end), and Monetization Mastery (turn the skill into a side-business).
It also includes done-for-you templates you import in two clicks, over-the-shoulder training videos, and the same playbook 1,000+ students have used to save two hours a day. No coding required. If you can copy and paste, you can build this.
The podcast humaniser: common questions
Pulled from what readers and Automations Made Easy students ask most.
Do I need to be a developer to set this up?
No. You point the machine at your script repository and tell it how you talk, then let it rewrite for the ear and strip the AI tells on its own. The skills you need are knowing your own voice and knowing what a natural episode sounds like, which is exactly what Automations Made Easy teaches. The cleaning is handled for you.
Why rewrite the script instead of just recording it?
Because AI writes for the page, not the ear. People do not write the way they speak, so a script that reads perfectly still sounds stiff when a voice reads it out loud. The rewrite reshapes the words the way a person actually talks, and that is the difference between a listener staying to the end and dropping off early.
How does it know which phrases scream AI?
AI keeps writing the same way, reaching for the same openings, transitions, and words that we have all learned to spot. The machine hunts those known tells and replaces each one with plain, human wording. I keep the method light here on purpose, but the result is what matters: nothing in the finished episode makes a listener think a machine wrote it.
Will people be able to tell it is a cloned voice?
Not once the script is clean. The reason a cloned voice sounds off is usually the words, not the voice, because AI phrasing does not fit natural speech. Once the script is rewritten for the ear and cleared of tells, the voice has natural words to read, so the episode sounds like a real person. Nobody listening suspects a thing.
Can I offer this as a service to other people?
Yes, and it is a solid income stream. Most people using AI to write scripts end up with something that sounds like basic ChatGPT, and you can take those scripts and make them sound a million times better for a monthly fee. It is recurring work, because they keep producing scripts, and the numbers on this page stay modest on purpose, but the upside is real.
More Money Makers like this one
Built from the same handful of mechanics. Each one captures, converts, or earns money that would have walked out the door.
Build this podcast humaniser yourself, or learn the mechanics inside Automations Made Easy.
If you want to learn the mechanics behind content automations like this and build your own at home, Automations Made Easy is the playbook. Step by step, no code, plain English. If you want to talk through how to clean your scripts, strip the AI tells, and turn a stiff library into episodes that sound human and sell, I take a small number of consulting clients each month.
€497 one-time · Lifetime access · 1,000+ students
