#04 - The Voice Memo Method.
How a four-minute recording becomes a 178-word post that doesn't sound like AI wrote it. The full system, four steps.
Last Wednesday a founder I work with sent me a four-minute voice memo. By Friday she had a LinkedIn post with 80 reposts, three inbound enquiries, and one comment that read “this is the first thing I’ve read this year that didn’t feel like AI wrote it.”
The memo was 4 minutes 12 seconds. The post was 178 words. The work between memo and post was 25 minutes of mine.
The system below is what I did in those 25 minutes. I run it with every new client. I run it on myself every Sunday. It’s the full playbook. Four steps. Repeatable. AI-optional. The whole thing fits on a postcard.
I call it the Voice Memo Method. You don’t need to call it anything. You just need to run it.
Step 1 · The Memo
You don’t sit down to “record content.” You sit down to record what’s actually in your head right now.
The prompt is one of three:
What’s something at work this week that pissed you off, made you laugh, or made you change your mind?
What’s something you said to a customer or colleague in the last 48 hours that surprised them?
What’s the thing about your industry that you’d say at a dinner party but never put on LinkedIn?
Hit record. Talk for four minutes. Don’t structure it. Don’t introduce yourself. Don’t pre-write. Specifically: don’t think about who’s listening.
The reason this works is the filter framework from two weeks ago. Three filters strip the voice out at the translation layer. The voice memo sits upstream of the filters. You haven’t started translating yet.
If you stop and start, restart. You’re allowed three takes. After three, ship whatever you’ve got. The third take is usually the truest one anyway.
Step 2 · The Transcript
Run the memo through any decent transcription tool. Whisper. Otter. Wispr Flow. The native iPhone transcript will do. Doesn’t matter which.
What you get back is messy. False starts, “um”, repeated half-sentences, the moment you went off on a tangent and came back. Don’t clean it up. Read it twice.
You’re looking for two things, both of which will jump off the page if you’re paying attention:
The sentence you said with energy that you’d never write down on purpose.
The cause and effect you snuck in without realising.
Mark them. That’s the surgery.
If nothing jumps, the memo prompt was too soft. Re-record with the harder version. The third one in the list above is the most reliable for surfacing real material because the dinner-party frame removes the LinkedIn brain.
Step 3 · The Spine
Every post that lands has one cause and effect sentence. Not a hook. Not an opinion. Cause and effect.
Cause: the freelancer smoothed your edges.
Effect: your sentence is now generic.
Cause: the AI tool was trained on the median.
Effect: it can only give you the median back.
Cause: you’ve been rewarded for not being weird in public for fifteen years.
Effect: you self-translate before you even open the doc.
Look at your transcript again. Find the cause and effect that’s already there. Almost always it’s one sentence you said quickly, possibly under your breath, possibly as an aside. Not the obvious “main point.” The thing you said in passing because it felt obvious to you.
That sentence is the spine of the post. Everything else hangs off it.
If you can’t find one, the post isn’t ready. Two outcomes from here. Re-record with a more specific prompt. Or shelve and move on. Don’t force a spine that isn’t there. Forced spines make forgettable posts.
Step 4 · The Rewrite
Now you write the post.
The rule: every sentence either supports the spine or earns its own cause and effect.
What gets cut:
Hedging phrases (”I think”, “in my view”, “personally”)
Throat-clearing openers (”So I’ve been thinking about…”)
Generic transitions (”here’s the thing”, “let’s dive in”)
Anything you wrote because you thought it sounded “professional”
What stays:
The specific number, name, or moment
The opinion someone could disagree with
The cause and effect
Order matters more than people think. The spine sits at the bottom of the post, not the top. The hook is whatever sentence is most likely to make a stranger keep reading. The middle is the proof. The end is the spine, restated cleanly. Posts read like an argument with the punchline in the right place. Punchline first kills the post.
For the founder I worked with last Wednesday, the spine was: most of our customers don’t care about the platform; they care that we took the part they hated and made it boring.
Hook (top of post): “Most of our customers don’t care about our product.”
Middle: three sentences of context, one specific customer story, one number.
End: the spine, restated. “They care that we took the part they hated and made it boring. That’s what we sell.”
178 words. 80 reposts. 25 minutes of work.
Why this works when “write better hooks” doesn’t
Most founders try to write better LinkedIn posts. That’s the wrong framing. You can’t write your way out of a translation problem. The translation problem is upstream of the writing.
What you can do is move further upstream than the post. Record before you translate. Find the spine in the recording. Cut the filter. Ship.
The Voice Memo Method is one of five extraction systems I run with paying clients. Next month’s paid post unpacks the second one, the system I use when the client doesn’t have time to record memos but has 90 minutes a week of internal Slack messages and Loom videos. Different surface, same principle.
For now, run this on the next thing you’d ordinarily ghostwrite to yourself. Then send it. See what happens.
That’s the work.
Sarra The Ghost
If you want me to run this on a memo you record, that’s what a Voice Audit is.
DM me for deets.