"I'm writing this entire article by speaking." A week after switching from Aqua Voice to Superwhisper × Scribe

To start with a confession, this paragraph you are reading right now is entirely written by me speaking (strictly speaking, I've edited what I wrote).

I've barely touched the keyboard. While my cat sleeps next to me, I'm just muttering to myself. Yet, Japanese text—and pretty decent Japanese at that—appears on the screen.

If this were me a week ago, I would have looked at this skeptically, thinking, "Yeah, that's the kind of thing that only happens in demo videos."

That's because, until a week ago, I was using an app called Aqua Voice.

The story of how my "excitement" for Aqua Voice faded

Aqua Voice isn't bad. In fact, when I first used it, I thought, "Wow, this is amazing." It picks up Japanese reasonably well and even adds punctuation thoughtfully.

However, as I kept using it, the excitement gradually faded.

This isn't so much a technical complaint as it is an emotional one. The feeling of "Oh, it's that same word mistake again" or "No, that's not where you should break the sentence" accumulated little by little every day, until one day I landed on the conclusion that "it's faster to just type it myself."

I think voice input runs more on a reserve of trust than on accuracy. The moment you stop believing that "this thing will pick up what I say," you stop using it.

I think that's where I had arrived.

What does it mean to be able to choose "Scribe" within Superwhisper?

That was when I encountered Superwhisper.

Superwhisper is a voice input app, but it's a bit unusual. You can choose the "transcription engine" used inside it yourself.

There are options lined up like "Use OpenAI's Whisper" or "Use ElevenLabs' Scribe," and you can set whichever one you like.

I chose Scribe. The reason is simple: I was already using Scribe for other purposes in my English studies, and I knew through other channels that "this thing is strong with Japanese."

In other words, my setup is Superwhisper as the shell, with Scribe as the engine.

The moment I switched to this, I had the feeling that everything I had almost given up on with Aqua Voice had returned.

The "spaces between Japanese words" problem that stumped me in the first 30 minutes

However, the first 30 minutes were quite hellish.

On the day I installed it, when I was speaking in front of my Mac as usual, the text that came out was like this.

There are spaces between every single word, all of them have spaces.

In English, that's correct. But when this happens in Japanese, the display breaks everywhere the moment you copy and paste it. It looks weird when sent on LINE, and it looks weird when pasted into a manuscript.

"Wait, what was the point of turning on Scribe..." I thought for a second.

Conclusion: Turn Realtime OFF for Japanese

The correct answer was in the Superwhisper settings.

Turn Realtime (real-time transcription) OFF. That's it.

When you turn Realtime on, words appear as you speak them. This feels great in English, but in Japanese, it was designed to display with spaces between every word (at least in my environment).

When you turn it OFF, the text appears all at once after you finish speaking. There is a "wait" of about a second, but in exchange, the Japanese that appears is without spaces and with reasonably correct punctuation form.

For me, this finally made it usable as a "replacement for Aqua Voice" on a daily basis.

The setting is located within the Superwhisper preferences, inside the detailed settings for each Mode. My current workflow is to create one mode for Japanese and keep only Realtime turned off there.

The moment you realize you're "not typing"

Since switching to Superwhisper × Scribe, a small change has occurred in my work.

For example, when I ask Claude Code to "write a script like this." Previously, I would type on the keyboard while organizing the sentences in my head before hitting enter.

Now, I just speak exactly what I'm thinking.

Like, "Uh, um, this, first I want to sort by file name, and then I want to take only the latest one from there," in that exact tone.

Even so, Claude Code understands my intent perfectly. In fact, there is sometimes more information than when I type it out neatly (perhaps because the "uh, um" parts contain information about what I'm struggling with).

When I realized this, I looked at my hands and realized I wasn't touching the keyboard.

I'm not really typing, yet the work is getting done. This feeling was actually something I didn't have during my Aqua Voice days.

If you had reached the point of "I'm done with voice input"

I am someone who had pretty much given up on voice input. I'm the type who would say, "The Japanese accuracy just isn't there," and end up going back to the keyboard, over and over again.

And yet, for the past week, I have been writing by speaking for a significant portion of my writing.

As I wrote at the beginning, this article you are reading right now is almost entirely Japanese that came out of my mouth (I had a little help from AI, but I only actually typed a very small part of it).

If you have tried native macOS voice input or Aqua Voice in the past and stopped at "Well, I guess this is as good as it gets," I want you to try just these two things: selecting Scribe within Superwhisper + setting Japanese mode to Realtime OFF.

Just like me, you will probably have a moment where you think, "Oh, I'm not really typing, am I?"

When that happens, think about it once more.

Of the time you usually spend writing, what percentage was actually spent on "writing the text"?
Isn't the rest just the transportation cost of getting what's in your head out through the keyboard?

I've realized once again that for me, that was quite often the case.

#VoiceInput #Superwhisper #Mac #WritingEnvironment #JapaneseInput