Turn a voice note into a LinkedIn post, not a cleaned-up transcript
Talk for two minutes about the thing you actually did this week. Get four posts back, each making a different argument from the same note. Every number in them is one you said out loud.
“We got onboarding from nine days down to two, and honestly the thing that did it wasn’t the automation…”
- Tactical Authority
The three things we cut to get onboarding from nine days to two
- Vulnerable Community Builder
We spent a year blaming the customer for a problem we built
- Strategic Executive
Slow onboarding is never an onboarding problem
- System Builder
The handoff checklist that took our onboarding from nine days to two
One note. Not four openings on the same idea, four different points.
You have had the thought. You had it out loud, on a call, in the car, halfway through explaining something to someone who asked. It was sharp when you said it. Then you sat down to write it up, produced four paragraphs of LinkedIn voice that sounded like everyone else, and closed the tab.
Speaking is not a faster way to write. It is a more honest one.
Something happens when you talk about your work instead of typing about it. You stop reaching for the phrase you think a post is supposed to contain. You use your own examples, in your own order, with the numbers you actually remember. That version is better than the one you would have typed, and it is the version nobody ever sees.
The obvious fix is transcription, and it does not work. A transcript of you talking is not writing. It has your hedges in it, your restarts, the sentence you abandoned halfway. Clean it up and you get something readable that still reads like a recording, because “kind of” and “sort of” are the residue of speech, not of thought.
PublishFlow does not tidy the transcript. It reads it for what you were actually claiming, then writes that as prose, keeping your examples and your turns of phrase and dropping the parts that were only there because you were talking. And because it can see nothing except what you said, it cannot put a number in the post that you did not say out loud.
How it works
Say it. Up to five minutes.
Press record and talk the way you would explain it to one person who asked. No prompt to fill in, no topic to pick. When you stop you can play it back and record it again before anything is sent.
Read the transcript. Fix the one word it got wrong.
It comes back editable, not final. Correct a name or a figure and that becomes the source of truth for everything written after it.
Get four posts, and pick the argument.
The same note written four ways. You are not choosing between four openings on one idea. You are choosing what your point was, after you have heard yourself make it.
Tactical Authority
Practical teaching. "Here is what I learned."
Vulnerable Community Builder
Personal story. "Here is what it taught me."
Strategic Executive
Pattern-spotting. "Here is what this means."
System Builder
Framework. "Here is how to replicate it."
The microphone is not only on this page
Anywhere PublishFlow needs your actual opinion rather than a source document, you can speak it instead of typing it. Voice is an input, not a feature.
- A note that becomes posts on its own
- Your take on an article or case study you read
- Your angle on a post you are about to react to
- Your reaction in the comment composer
A real note, and the post it became
Both halves below are verbatim from our own account. Nothing has been tidied for the screenshot, which is rather the point: read the left side and you can hear someone talking.
A lot of companies think that SEO in 2026 is still all about writing blog posts on your website. While blog posts is still a good, like publishing blog posts on your website to rank on Google is a good strategy, but now SEO has evolved beyond just writing blog posts. So especially now that AI search is on the rise, so because most people are not just Googling and waiting for Google to tell them what to do, like what to do, like to tell them the solution to pick, they are asking AI tools for recommendations.
Three restarts, two abandoned clauses and three instances of “like”. This is what everybody actually sounds like, and it is why a tidied transcript still reads like a recording.
Most companies running SEO in 2026 are still optimizing for a user who barely exists anymore.
The blog post strategy isn’t dead. But if it’s your entire SEO strategy, you’re only showing up in one place, and that place is losing traffic to AI tools fast.
Here’s what I keep seeing with clients: they’ve been consistent with publishing, good topics, decent rankings, solid traffic from Google. Then they ask why they’re not getting recommended when someone asks ChatGPT or Perplexity which tool to use in their space.
The answer is almost always the same. They’ve built visibility on their own website but nowhere else.
Same claim, same examples, same person. Not one spoken sentence carried across intact.
The rule it writes under
A voice note is treated as spoken language, and the writer is instructed never to transplant a spoken sentence into the post as it stands. Hedges and filler are named as transcript residue, not writing. That is the line in the product that separates this from tidying up a recording.
It cannot invent a number
Before anything is written, your transcript becomes a ledger of the facts it contains, and every figure in the finished post is checked back against it. Say you went from nine days to two and the post says nine days to two. Give no number and the post has none.
Four arguments, written in parallel
Not four drafts of one post. Four different cases the same material can make, generated at the same time so none of them is a revision of another. Choosing between them is the fastest way to find out what you actually meant.
The mockups elsewhere on this page are examples, not customer data. Voice notes are a young part of the product and we would rather show you the one real thing we have and how the writing is constrained than a number we made up.
What it will not do
The limits, stated plainly, so you can decide before you sign up.
It is not a dictation tool.
It will not type your words into LinkedIn’s own composer. LinkedIn has no voice typing there, and this does not add one. If you want your exact words transcribed verbatim into a text box, this will frustrate you, because it rewrites on purpose.
It is not a voice notes app.
No library, no folders, no search across your recordings. A note becomes posts and then it is finished.
Nothing keeps the audio.
The recording goes straight to transcription and is never stored. You cannot play it back tomorrow. That is a deliberate trade, and it means there is no archive of your voice to leak.
Five minutes per note.
The countdown is on screen and it stops on its own. Longer than that is a meeting, and there is a different tool for meetings.
There is no mobile app.
Recording works in your phone’s browser, but you cannot share a memo into PublishFlow from your phone’s Voice Memos.
Record here, do not upload.
You cannot drop an existing audio file into this flow. If the audio is a call, the notetaker is the tool for that.
Works with
LinkedIn post generator
What happens to the post you picked, including the reader test on your first line.
Meeting notetaker
For the things you said out loud on a call, rather than into a microphone on purpose.
The content engine
A voice note is one input. Monday arrives drafted from all of them.
Voice profiles
The reason the four posts sound like you rather than like a tool.
Questions
Can I record a voice note on my phone?+
Yes, in your phone’s browser. There is no PublishFlow app yet, and you cannot share a recording into it from your phone’s Voice Memos app. You record in the browser, in the moment.
Do you keep the audio?+
No. Your recording is transcribed and then it is gone. It is never written to storage. The transcript is what gets kept, and only as the source behind the posts you saved.
What if it mishears a word?+
You see the transcript before anything is written, in an editable box. Fix the word and it becomes the source of truth for all four posts. Names and numbers are the two things worth checking.
How long can a voice note be?+
Five minutes, with the countdown on screen. In practice two to four minutes is where the useful notes land. If you have more than five minutes of material, it is probably a meeting, and the notetaker handles those.
Does this work in Arabic?+
Yes. Set the recorder to Arabic and an Arabic note becomes an Arabic post, written in the register a professional actually posts in rather than translated stiffly. The source language decides the output language, so a mixed note follows whichever carries most of the material.
How is this different from a voice notes app that writes for me?+
Those tools give you back a cleaner version of your own talking, which is genuinely useful and is not a LinkedIn post. This one reads the note for the argument you were making, writes that argument four different ways, and refuses to put a number in the post that you did not say out loud.
Does a voice note use up my monthly posts?+
One note counts as one generation, even though it gives you four posts back. Full access includes fifty generations a month.
The best version of this was the time you said it out loud
Two minutes, once, about the thing you actually did this week.
Start freeFour posts from one note counts as a single generation