The AI Dictation App I Actually Use

I paid for Wispr Flow, tried OpenWhispr, and use Handy now. Free, local, and good enough that I stopped paying. Wispr Flow is still the best polish if the price fits. The rest of this post is why I landed there.

I was a fast typist who did not care about dictation

I type 60 to 90 words a minute, properly, no hunting and pecking. As a developer I type all day, so dictation never felt like something I needed. I knew it existed. Dragon was doing this back around 2006, macOS has had built-in dictation for years, and Raycast added an AI version. I ignored all of it, until someone talked me into trying Wispr Flow, and it changed how I work.

Why AI is what made dictation click

Old dictation typed exactly what came out of your mouth, filler words and false starts and all, so you spent as long cleaning it up as you saved. AI is the piece that fixed that. It drops the ums, punctuates, formats, and even sorts out who said what in a meeting. That is the jump from a raw transcript to something you can actually send.

One naming note, because it trips people up. The paid app is Wispr Flow, spelled without the h. The local tools further down run on Whisper, the open model from OpenAI, and one of them is even named OpenWhispr. Similar-sounding names, different things.

The best one I have used: Wispr Flow

Wispr Flow is the premium pick, and it earns that label. It cleans and formats as you speak, adapts the tone to the app you are in, learns your vocabulary, and works anywhere your cursor is, from Gmail to your code editor. Speaking, I move around 150 to 200 words a minute, well past my typing. It has a Notetaker for meetings now too, with speakers named and a summary at the end.

Here is the honest part. When I used it about two months ago, Pro ran about $20 a month, and I could not justify that for dictation alone. The free plan gives you 2,000 words a week with no card, so you can feel the difference before deciding. If you want that polished premium experience and the price fits your work, Wispr Flow is the one I would put money on. It just stopped being worth it for me specifically.

The developer trick worth knowing

Wispr Flow pairs well with an IDE like Cursor. In Cursor you reference files and directories with an @ sign, and with Flow I could just say “tag” and the file name out loud to pull that reference in. Files and directories, at least when I used it, not skills, so check their docs for the current state.

I do not miss it, and here is why. AI has gotten fast enough that the workaround is easy. If I keep my context tight and stay on one thing at a time, I am not confusing the model, so there is nothing to tag. And since I am talking anyway, saying “use the skill called X” out loud is just as quick as a dedicated tag. The feature was great. I just do not reach for it now.

Going free and local: OpenWhispr

Price is what pushed me to look, and open source had an answer. OpenWhispr bills itself, in its own words, as the free, open alternative to Wispr Flow. It runs on Mac, Windows, or Linux, keeps everything local and private with no telemetry, lets you pick your speech model or send audio to the cloud for speed, and handles dictation, meeting transcription, and notes. It is a lot of tool for zero dollars.

It was not for me, for two reasons. Accuracy on a local model was a step down from a paid cloud service, which is the tradeoff for keeping everything on your device. And it was heavy: on my Mac mini M4 with 24GB of memory, running a local model during a meeting, on top of my dev tools and a video call, was more than the machine had, and there was a cold start when the model unloaded between uses. If your machine has the headroom, it is well worth a look.

What I use now: Handy

Handy (handy.computer) is where I landed. It is open source and local like OpenWhispr, but it only does voice, which is exactly what I wanted. It keeps the model warm for up to fifteen minutes, so there is no cold start, and it is quick at both capturing and processing. It is not as accurate as Wispr Flow, but it is accurate enough that I do not fuss with it. It is free, and because it is open source I can fork it and shape it toward the same kind of polish Wispr Flow offers, without paying monthly.

Most people do not have the time or the technical appetite for that. If you want the premium features without building anything, Wispr Flow is the better fit. I am putting this out there as what works for me, not as a rule for everyone.

Which one is for you

If you want the most polished, most capable dictation and the price works, start with Wispr Flow. Try the free 2,000 words a week and see if it clicks. If you want free, local, and private, and good enough is good enough, a tool like Handy or OpenWhispr will surprise you. That is the trade, and after living on both sides of it, free is where I am staying for now.

---

Some links here are affiliate links. If you start a paid plan, Artech may earn a small commission at no extra cost to you. I moved to a free tool myself and was not paid for this; the opinions are my own.

Run the business while I manage the tech.