Wispr Flow: Middle Mouse Button will make you work faster

01/  The short version

Wispr Flow is dictation that sits on top of whatever you are already typing into. Hold a button, talk, let go, and the text lands in the active window with the filler words and the false starts cleaned out. I get up to 170 words per minute out of it, and I use it for most of what I write in a day: Slack, Linear issues, documentation, and the long briefs I hand to AI agents.

Two pieces of hardware are what took it from a novelty to the thing I reach for. Get a mouse with a scroll wheel you can press, and a headset with a dedicated microphone. Bind dictation to the scroll wheel click, talk into the headset, and the rest of this is why those two and what changed once they were in place.

02/  Bind it to your mouse

Flow ships bound to Fn on a Mac and Ctrl+Win on Windows. Both work fine. If you dictate into one window all day you may never want anything else.

My day does not look like that. I am clicking into a Slack thread, then a Linear description, then a terminal, then the prompt box of whatever agent is running. The mouse is how I get to every one of those boxes. With the trigger on middle click, pressing the scroll wheel down, the click that puts my cursor where I want it and the button that starts dictation are the same hand in the same place, and I start talking without moving.

That is the whole gain, and it compounds across a day of short bursts into a lot of different places. After about a week I stopped noticing I was pressing it, which is the point where a tool disappears and you just use it.

Flow is specific about which buttons it accepts. Middle click and mouse buttons four through ten work, alone or behind a keyboard modifier. Left and right click are out. So is the Apple Magic Mouse, so are built-in trackpads, and so is anything with three buttons or fewer. On a bare laptop you are stuck with a key.

03/  Then fix the microphone

I went through three before it was right, and the order is worth knowing because the middle one looks like a solution.

The MacBook microphone sits up by the hinge and hears the entire room. Your distance from it changes every time you shift in your chair, so the input level moves around while you talk, and it picks up your own keystrokes.

AirPods were the obvious next try and they were not it either. Bluetooth drops to a much lower audio bandwidth the moment the microphone opens, so the same earbuds that sound excellent playing music are a downgrade the second you talk into them.

A wired headset with a boom microphone solved it. The mic holds one distance from your mouth whether you are leaning back or hunched over a spreadsheet, and it sits close enough that the room stops registering. With that on, Flow is basically flawless for me.

The mouse gets you using it. The microphone makes it worth using.

04/  What I get out of it

The first thing that changed was length. Background that used to stay in my head because typing it out was a chore now goes into the prompt. Agent briefs are the clearest case: they want the contract terms, the history, and the reason this particular account is strange, and every one of those sentences used to cost something to produce. At talking speed you just say all of it.

The second is where a first draft comes from. An empty box is a hard place to start. Talking gets a rough, badly organized version of the whole thing out in one pass, and editing a rough version is a different job from writing one. For a Linear issue I say all of it badly and then clean it up, and by the time I start editing the structure is already there.

The rest is small and adds up. It pastes into whatever window has focus, so Slack, a Linear description, a terminal and an agent prompt all work the same way. It strips filler and keeps the correction when you talk over yourself mid-sentence. It learns the words you fix, so after a few weeks it stops mangling your product names and your colleagues.

05/  What it noticed about me

Flow keeps an Insights page in the desktop app. One tab counts the measurable things, and another builds a Voice Profile out of how you actually talk. That one is the fun part.

Mine currently reads Structure Refiner, described like this: voice is your go-to tool for refining complex structures, whether it is reworking Salesforce workflows or clarifying project details. Your dictations often focus on streamlining processes and ensuring every detail aligns with broader goals.

• Catchphrase: I think we can
• Most used word: should
• Most corrected word: and
• Peak time and place: Thursday at 2pm, in Slack