Turn a meeting, an article, or a case study into one post you would actually publish
Paste the link, drop the transcript, or just talk. PublishFlow reads your material for the one thing only you could have said, and writes that.
No credit card
https://example.com/research/why-migrations-stall
Link detected. We’ll fetch the article for you.
One input, whatever you brought. No format picker, no source-type grid.
You are not short of material. There was the client call where you finally explained the thing properly. The article you sent to four people this week. The case study nobody outside the company has ever read.
It all sits there. And on Tuesday you open LinkedIn and still have nothing to say.
Reformatting was never the hard part
Every tool in this category means the same thing by turning one thing into another: reformatting. One webinar into fifteen clips. One blog post into a carousel, an infographic and a thread. The pitch is always a multiplier, and the multiplier is always the point.
That solves a problem you do not have. You are not short of posts. You are short of one specific thing you know that the people reading you do not, said in a way that could only have come from you.
Reformatting cannot produce that, because the interesting part was never the format. It was the moment the client went quiet. The number that did not go the way you expected. The sentence in that article the author was clearly avoiding.
So PublishFlow does not reshape your material. It reads it, the way a good ghostwriter reads it, looking for the thing you would have told someone over coffee. Then it writes that one post.
One source. Not fifteen posts. One worth publishing.
How it works
Put it in the box.
There is one box. Paste a link and it gets fetched. Paste a transcript and it is recognised as one. Paste anything else and it is treated as your own material. Or upload the .vtt straight off the call, or press record and talk for two minutes instead.
Pick the moment, not the framing.
It reads the whole source and comes back with what is actually in it. If you were there, that means moments: what happened, the words that were said, the part that was contested or conceded or expensive. You choose which one is worth a post.
It writes that one post, in your voice.
Written against your voice profile, your audience and your offer. Every factual claim has to trace back to something in the source you gave it, and the post is checked and repaired before you ever see it.
The detection is a guess, and it is correctable. The next screen shows you what it decided your source was, so a wrong guess costs one click rather than a bad post.
What comes back from one implementation call
Example- 1
They ran the old process in parallel for four months
We never turned the old one off. Nobody wanted to be the one who did.
A specific decision, a quote nobody else on LinkedIn has, and a claim people in your field will argue with.
- 2
The migration finished in week two. The trust took until month five.
The same argument from further out. No quote behind it, so it leans on you to tell it properly.
- 3
Nobody asked for a rollback. They just kept the old tab open.
The best line in the call, and the thinnest evidence under it.
Rests on a private conversation
You pick one. One post gets written from it, not five variations of it.
Where the material came from changes what it looks for
Not the tone. The job. Reading your own call and reading somebody else’s article are two different tasks, so they run as two different tasks.
You were there
Your call, your client work, your company story, a post you wrote, a voice note.
It mines moments, because you have something to protect: the thing that actually happened. You get the three most worthwhile ones, and one post is written from the one you choose. Three options about three different things, not three framings of one.
You were not there
An article somebody sent you, a case study you found, a report you disagreed with.
It hunts readings instead, because you have no moment of your own here, and asks what your take is. Your take becomes the claim the post argues, and the material you found becomes the evidence for it rather than the subject of it.
What stops it drifting away from your source
There is no customer result to show you here yet, so here is the mechanism instead. The mechanism is the claim.
It quotes, it does not paraphrase
Extraction pulls verbatim phrases out of your source and the writer builds from those. So the texture in the finished post is your material and your phrasing, rather than the model’s vocabulary smoothed over the top of it.
Every number is checked against the source
A ledger of the facts is built from what you supplied, and the finished post is checked against it in code, with no extra model call. A figure that is not in your material does not survive into the post.
The writer sees the window the ideas came from
Both are thirty thousand characters, and they are equal on purpose. An idea mined from minute forty of a call reaches a writer that can still see minute forty, so it never has to fill the gap by guessing.
“5 lessons from our latest implementation: 1. Communication is key. 2. Change management matters. 3. Get buy-in early…”
Every sentence in the call, flattened into the shape of a post. Nothing in it could only have come from this call, which is what makes it interchangeable with everyone else’s.
They ran the old process in parallel for four months, and nobody would say why.
We never turned the old one off. Nobody wanted to be the one who did.
One of these contains a sentence that exists in exactly one transcript in the world, and you are the only person holding it.
One honest note about what gets used
The article and link path is the most used input in PublishFlow today, ahead of transcripts, case studies and voice notes. We are telling you that without a percentage attached, because the number of people behind it is small and a percentage would dress a direction up as a statistic. Every moment, quote and post fragment on this page is an example rather than customer data.
What it will not do
The limits, stated plainly, so you can decide before you sign up.
It will not cut your video.
No YouTube link, no MP4, no clips and no captions. If your source is a recording, bring the transcript and it works from that.
It will not read a podcast.
There is no audio file upload. The only audio it takes is you talking into it for a couple of minutes.
No PDFs and no Word documents.
Paste the text. Plain text, Markdown, .vtt and .srt are the files it takes.
It cannot fetch a LinkedIn URL.
LinkedIn blocks it, and it will tell you so rather than returning something thin. Paste the text of the post instead and everything else works normally.
It cannot get behind a login or a hard paywall.
And it gives up after ten seconds rather than hanging. Paste the text when a page will not open.
It does not read forever.
Sources are trimmed at about thirty thousand characters, roughly a two hour transcript. Longer than that and you should bring the part that matters.
It makes no carousels and no images.
What you get is the text of a post. Nothing in the product generates a carousel today.
LinkedIn only.
There are no X, Instagram or multi-network variants, and none are planned.
It will not give you fifteen posts from one source.
That is a decision rather than a missing feature. You get up to three, each about a different thing in the material.
Works with
Meeting notetaker
Send it to the call and skip the transcript step entirely.
The content engine
Connect your sources once and the drafts arrive without you pasting anything.
Voice notes
For when the material is in your head rather than in a file.
LinkedIn post generator
What happens after you pick one, and why the result sounds like you.
Questions
What can I actually paste in?+
A link to an article or a case study, a meeting transcript, or any text you have written. You can also upload a .txt, .vtt, .srt or .md file straight off a call, or record a voice note in the browser. There is one box for all of it and PublishFlow works out which is which.
Do I have to tell it what kind of source it is?+
No. It detects a link, a transcript or plain material on its own. If it guesses wrong, the next screen lets you correct it in one click, and that changes how the post is written rather than just the wording.
Why do I only get one post from a call, when other tools give me fifteen?+
Because a moment survives being told once. Generating fifteen versions of one thing that happened turns it back into a generic framework, which is the failure everybody recognises in AI posts. When you were present for the material, PublishFlow finds the three most worthwhile moments in it and writes one post about each, so your three options are about different things rather than three framings of the same thing.
Can it turn a YouTube video or a podcast into posts?+
Not today. It takes text and transcripts, and it takes you talking into it. If you have a recording, bring the transcript and it will work from that.
Can it read a link to a LinkedIn post?+
No, LinkedIn blocks that. Paste the text of the post instead and everything else works the same way.
What happens to a very long transcript?+
Sources are trimmed to about thirty thousand characters, roughly a two hour call, and the same window is given to both the part that finds the ideas and the part that writes the post. Nothing is mined from material the writer cannot see.
You already did the work worth writing about
You already had the conversation. You already read the article. Fourteen days free, no credit card. Paste one thing and see what it finds.
Start freeA link, a transcript and a voice note all draw on the same monthly posts