From audio to document: a walk through what you actually get

So you’re weighing up Whistle Enterprise, and there’s really only one question that matters: what do you actually get out the other end? It’s the right thing to ask. All the privacy talk in the world doesn’t help if the document that comes out isn’t one you’d actually use. So let me walk you through exactly what lands on your machine when a meeting goes in…

The three things you end up with

Run a meeting through Whistle Enterprise and you’re left with three things, all sitting in a workspace on your own computer:

1. The recording - the original audio. Either something you recorded live, or a file you already had, audio or video (it pulls the audio track out either way).
2. The transcript - the recording turned into text. It works out the language on its own, you don’t have to tell it, and it can split the transcript up by speaker so you can see who was talking when.
3. The document - the written up version of the meeting. This is the bit our own fine tuned AI model produces, the one we built for writing up meetings and made to run on a normal laptop CPU. It reads the transcript and writes the document right there on your machine. Nothing gets sent off anywhere.

Now those three are kept deliberately separate, and that matters more than it sounds. The recording is the source of truth, it’s what actually happened. The transcript is what the model reads. The document is the thing you’d hand to a colleague. Keeping them apart is what lets you check the document against the transcript, and I’ll come back to that because it’s my favourite part.

What’s actually in the document

This is the question I get asked most, so let me be straight about it. The document isn’t a one paragraph summary, and it isn’t the transcript with the ums taken out. It sits in the middle.

The model reads the transcript and writes it up into sections, each with a short lead in and bullet points where something is genuinely a list. It builds those sections around what the meeting actually covered rather than forcing the same template onto everything, so a long technical discussion and a quick stand up come out looking pretty different, because they were different.

On top of that there’s a separate pass that pulls out action items, the genuine follow ups where someone actually said they’d do something. It’s deliberately strict: a line that just describes how things are, or something that already happened, doesn’t get called an action. And it only puts a name against an item when the meeting named one, it won’t invent who’s on the hook. If nothing in the meeting was a real action, you don’t get a padded list of busywork, you get nothing, which is the honest answer.

Now let me be straight about what it doesn’t do, because this is exactly the kind of thing you’d find out the hard way otherwise. It does not go through and tag every line with who said it. The speaker labels live on the transcript, not in the write up. So if your work needs “X proposed this, Y disagreed” baked into the document itself, that isn’t what comes out, and it’s better you know that now than after a meeting that mattered.

Checking it against the transcript

Here’s my favourite part, and the bit I think actually earns its keep. Highlight a sentence in the document and Whistle Enterprise points you to the part of the transcript it came from. Go the other way, highlight something in the transcript, and it shows you where it ended up in the write up.

Let me be straight about how that works, because it’s easy to oversell. It matches on the wording, the specific words and names and numbers two sentences share, so it’s a strong best guess at “this came from around here”, rather than a guaranteed line by line mapping. In practice that’s exactly what you want when someone says “hang on, where did that come from”, you highlight the line and jump to the bit of conversation behind it. And the recording is still sitting on your machine, so listening back to that moment is the next step, and that’s entirely in your hands.

Getting it back out

When you’re ready to share, there are three ways out: PDF, Word (.docx) and Markdown. The PDF and Word exports come with three themes, Standard, Formatted and Professional, so the same write up can go out plain for an internal share or a bit smarter for a client without you reformatting a thing. Markdown just hands you the raw text to drop into wherever you keep your records.

You can fold the transcript into the export too, and that’s where the speaker labels turn up, as a transcript section underneath the write up. The original recording stays put on your machine, the export is only ever a copy of the document as it stands at that moment.

Try it on a meeting you’ve already got

If that’s the shape of thing your work needs, the trial is the obvious next step, and it’s the full product for 30 days, no card and no account. The best test isn’t a fresh demo meeting, it’s a real recording you already have sitting on your drive. Run that through and read what comes out, that’ll tell you more in ten minutes than any marketing page.

The licence page has the pricing and the FAQ, the security notes cover what the app does and doesn’t do at a system level, and the free 30 day trial is on the product page.