Published on

Narrated, then Tidied by AI

Narrated, then Tidied by AI

by Timothy Coleman - 30/May/2025

Why Narration Tops Typing

I have a lot of ideas. Some are half-baked, some turn into full-blown articles, and many go nowhere. Getting them out of my head quickly used to mean hammering away at a keyboard. Touch-typing with my eyes closed felt impressive at university, yet even that is glacial compared with simply speaking. Modern voice tools claim they can triple your output; in my experience, they are not exaggerating.

SuperWhisper: My Current Favourite

My go-to app is SuperWhisper. It sits behind a keyboard shortcut, pops up like Spotlight, and starts transcribing the moment I talk. I keep an offline high-quality model for long-form writing such as this post, but switch to the ultra-fast model for quick tasks like WhatsApp replies. Offline processing keeps my words on my laptop, cuts cloud costs, and spares a data centre’s GPUs.

A Quick Comparison

  • Otter and SwiftKey Voice: decent for meetings or phones, but nowhere near SuperWhisper’s accuracy.
  • System voices such as Siri, Google Home, or Alexa have fallen behind in natural interaction.
  • ChatGPT’s voice mode feels like a phone call with a real character, but again requires cloud round-trips.

Workflow Tweaks That Matter

  1. One tap to record – I have mapped the SuperWhisper shortcut to my trackpad via BetterTouchTool.
  2. Model choice by task – HQ for articles, fast for chats.
  3. Immediate tidy-up – once the raw transcript is ready, I feed it to an AI editor (like ChatGPT) for structure and clarity while preserving my voice.

The result: clean copy in minutes rather than hours, without sacrificing ownership of the content.

Privacy and Efficiency

Running models locally echoes my approach to web apps: push as much work client-side as possible, keep server bills low, and cut energy usage at hyperscale. The same logic applies to voice. If my laptop can handle it in real time, why burn power in a data centre?

Narration Beyond the Laptop

Voice makes sense everywhere:

  • VR headsets – typing on a blurry Bluetooth keyboard is painful; narration would be friction-less, yet Meta’s software still lacks it.
  • Co-working spaces – I sometimes glance at fellow engineers chatting and wonder: is that a Zoom call or are they dictating code?
  • On-device AI – the SuperWhisper model is small enough that a headset or even a phone could handle it with ease.

Narration plus AI Cleanup: A Winning Duo

“AI writing” often implies letting a model draft entire articles. That is not my style. I speak my piece, ramble included, then rely on AI to reorder paragraphs, add headings, and remove fluff. The story stays mine; the polish comes from silicon.

Final Thoughts

Voice narration has changed my workflow more than vibe coding, code completion, or any IDE trick. It is faster, more natural, and—when paired with a good AI editor—produces clearer writing with less effort. If you have not tried narrating your next blog post, essay, or even daily notes, give it a go. You might never look back at the keyboard.