I've dictated over 120,000 words with my voice - these

What happened

Since February, I have dictated 120,896 words. In the last three months alone, I have dictated 50,234 words. I've done this across 2,314 individual microphone recording sessions, averaging about 24 words per dictation sequence. With improvements in AI, dictation quality, speed, and accuracy have come a long way. Over the years, I've tried to work with speech recognition many times, but it hasn't been very successful until quite recently.

Because of the regular need to do editing and superfine word crafting, I don't find voice dictation to be nearly as effective as hand typing on a keyboard like a caveman. On the other hand, I use voice dictation a lot when working with Claude Code or OpenAI's Codex to vibe code any of the products I'm working on.

I use it a lot in Slack and Google Chat when talking to my editors and some of my project partners. I use voice dictation quite a lot when replying to email messages. I use it a little bit less when composing new messages, but I use it there sometimes as well. I use it a lot when taking notes.

I also use it sometimes for search phrases in Google or prompts that I give to ChatGPT, Gemini, or Claude Code. ) I find that voice dictation is sometimes a little bit less precise and a little bit less disciplined than writing each word with the keyboard. But I can dictate at an average of 118 words per minute, while my typing is usually in the 70- to 80-word-per-minute range.

When I'm working on an article, I sweat each sentence, choosing the words and structure with great care. Typing lends itself to that kind of attention to detail.

But when I'm vibe coding, for example, and I'm discussing how I want a feature to be instantiated, how I want something to behave, or a bug that I've noticed, speaking is considerably faster and also considerably more gentle on my hands than typing it into the computer. I find two features to be mission-critical.

The first is a customizable dictionary, so that when I say something like ZDNET, the dictation product understands how I want it spelled and presented. The second is an on-the-fly correction capability, so that when I say something and then correct myself and re-say it, the version that lands in whatever I'm dictating into contains those corrections.

Why it matters

A typical article is roughly a thousand words. So if you look at it that way, I have dictated the equivalent of roughly 120 articles since February. That's not to say I don't type. I type a lot, but it's clear that I also use dictation a lot. I've adopted voice dictation as a primary input modality for two key reasons.

First, it helps to protect my wrist, which tends to have carpal tunnel symptoms. By dictating, I'm using my wrist a little bit less. In this article, I'll show you the three primary contenders that I've looked at and spent time with this year.

I'm also going to go over a number of also-rans and honorable mentions, just to give you an idea of the dictation tools that are out there and how they differ. When it comes to dictation products, there are a variety of factors to consider. Price is certainly one of them.

But there's also the question of whether or not you want your words sent up to the cloud to be processed or whether you want to keep them private and local. Accuracy is another factor to consider. Back in the day, it used to be much more of a problem.

Fortunately, the voice models have achieved a level of overall quality that allows basic English words to be transcribed quite accurately. Some of the dictation products support other languages, but since I work only in English (at least for human languages), that's what I'm covering in this article. You also need to be able to correct your words as you're dictating.

Some AIs go back and clean up mistakes you've made and provide you with clean text in whatever it is you're dictating into. I am dictating most of this article just as a proof of concept to make the point that an article can be dictated. But that's not the norm for me. Despite all the dictation I do, I don't usually dictate my articles.

I sometimes dictate a few lines of an article, but mostly I type them out. One of the reasons I don't use voice dictation for my articles is that voice dictation isn't particularly good at formatting.

It's quite good at just laying down text, but if you want to arrange something, create a table, create bullets in a certain format, have indented quotes, or add any of the physical artifacts of an article, you have to go back and do that by hand after you dictate.

What to watch

One oddly missing feature in all of the voice dictation products is the ability to go back and selectively edit already pasted-in words. Nothing can do this except the native Mac Voice Control or Windows Voice Access that comes with the operating systems themselves. Unfortunately, those two capabilities often conflict with the voice dictation products in inconvenient ways, so they don't combine well as integrated solutions.

Almost all the dictation products are initiated by a hotkey. I bind the dictation hotkey to a button on my mouse so that when I tap the button, the dictation starts or stops. This allows me to dictate regardless of what application or web page I'm in at the moment. It means I can do computer input even if my keyboard isn't in front of me.