Writing or Dictating?

What’s the Best Way to Generate Text (with a little help from AI)?

I am not very good at typing on a keyboard. Writing text this way is slow – but I have been used to it for a long time.

What about dictating the text and having an AI tool do the transcription (speech-to-text)?

Can I formulate my thoughts better by simply dictating them and letting the AI translate them into written words? Or is writing better because I can be more precise with my thoughts? When I write, I am used to pre-formulating sentences or thinking about them beforehand, but when I speak, I find it harder.

I asked ChatGPT about the difference:

Dictating text can sometimes be more difficult because it requires different cognitive processes than writing by hand. When writing, you can see the words on the page as you form them, giving you more control over the structure and phrasing of your thoughts. Dictating requires you to rely on your memory and your ability to articulate your thoughts in a coherent manner.

This may be true if I dictate first (to an audio file) and have it transcribed later into a text document using tools like Whisper AI or Otter AI (to name a few).

But there is another way : using speech-to-text tools like the ones built-in MacOS or Windows.

I use an Apple Mac that supports local speech recognition. It converts speech to text right on the computer (same for my iPhone) and can be used with most applications (word processor, email, etc.).

So I can see the text on the screen as I dictate. Similar to typing. But it is not the same cognitive experience.

Interesting finding: when I formulate a text in my head, it comes easily to me, and the texts often have a structure that I can’t create directly when I dictate (only through the editing process).

It is clear that dictation is a convenient way for many users (including me) to create text documents. It is also clear that dictation is a skill that needs to be developed. It does not come naturally the first time. You have to practice.

Dictated text needs to be edited just as much (or even more in the beginning) than text typed on a keyboard. Of course you can use AI tools here as well – but that is for another post.

In the future, we will communicate more and more with computers (or their future equivalents, just think of Apple Vision Pro) and programs by voice.

I will start practicing now.

(This text was dictated in German. I have used Deepl Write to improve the text for more clarity and DeepL Translate for the translation to English. ChatGPT was used for research and review. The final version war edited by me using a keyboard)