[Honest Review] I Tried 3 Voice-Input LLMs and Only Kept 2

Are you curious about voice-input LLMs, but stuck on the question, "Which one is actually good?"

I was the same way.
I would find one that looked promising, pay for it, realize "this isn't quite right," quit, and then try another one. Honestly, I spent months just in this loop.
I have finally settled down now.

In this article, I will honestly write about the three tools I actually paid for and used extensively, explaining "why I kept them" and "why I stopped using them."

What works for one person might not work for another, so I won't say "this is the correct answer."
However, the criteria I used to choose them should be helpful.

1. Conclusion first: The 2 I kept and the 1 I stopped using

My current voice-input environment looks like this.

• Typeless: When giving instructions to AI from a smartphone

• Aqua Voice: For PC work, and almost everything else

And Superwhisper is one I tried but have since removed.
It wasn't that the features were bad. It just didn't suit me.

From here on, I will write specifically about "why I kept" and "why I removed" each one.

2. Typeless: My partner for giving instructions to AI on my smartphone

What's great about Typeless is that it organizes what you say into a "clean instruction text" for you.

Imagine a situation where you are giving instructions to AI from your smartphone.
You want to quickly input something you thought of while walking while you're out.
But typing long sentences with a smartphone flick keyboard is exhausting. On the other hand, if you just talk to it messily, your instructions to the AI will end up a mess.

Typeless turns that cluttered spoken language into structured text.

For example, even if you ramble on like, "Um, I want to write the continuation of that article, and based on the flow of the previous one, make it a bit more casual...", it will output a properly organized instruction text. This is convenient.

It is especially helpful in situations like these.

• When you want to quickly throw instructions to AI while on the move

• When you don't have the energy to type long sentences but don't want to give messy instructions

• When you want to "get your thoughts into some kind of shape for now"

Conversely, it doesn't get much use when I'm working carefully on my PC.
Or rather, it tends to over-simplify things, and there aren't many situations where I need to input text that lacks my personal touch.

That's precisely why being able to commit to it as a "smartphone-only" tool makes it easier to use.
The paid version of Typeless is quite expensive, so I'm sticking to the free range, but it seems good enough for occasional use from my phone.

3. Aqua Voice: Why I quit once and came back

I will be honest about Aqua Voice.

I actually paid for it once, and then I quit.

When I first used it, my feeling was, "Well, it's not bad, but let's try others." Also, talking continuously is painful in itself, so when my enthusiasm for voice input cooled down, I canceled my subscription.
After that, I tried various methods including Superwhisper, and when I returned to Aqua Voice again, I realized, "Ah, this is the most stress-free one."

What's good about it is the accuracy of the text coming out exactly as I said it.

With voice input, if the conversion is even slightly off, the stress level spikes immediately. If "I didn't say that" happens two or three times, I just want to go back to the keyboard.
Also, it happens sometimes with Typeless too, but it's really stressful when it should have been input but doesn't output due to an error.

Aqua Voice has very little of that discrepancy. What I say becomes text almost exactly as is. It's fine even if I speak quite fast or mumble.
It's subtle, but this was the most important thing for daily use.

(By the way, there are parts I could only be sure about because I quit and came back. I'd like to think that the detour wasn't a waste.)

4. The story of why I quit Superwhisper (It wasn't a functional issue)

Regarding Superwhisper, I'll say it upfront: as a tool, it's totally fine. In fact, it has a good reputation among those who use it, and I didn't have any complaints about its functionality.

However, it didn't suit me. The reason is the "screen display."

Superwhisper has a waveform that moves quite a bit while recording. That visual movement bothered me.
I'm the type who wants to minimize the amount of information in my field of vision while working. When the waveform is wiggling around, my attention gets pulled toward it, and my train of thought is interrupted.

(Maybe this is an INTJ thing. When I enter focus mode, I become sensitive to visual noise.)

So, the reason I quit Superwhisper wasn't "lack of features" but "I couldn't focus."
It's purely a matter of compatibility. So, I think it's a good tool for people who aren't bothered by it.
Also, when using it from a smartphone, I want to turn it on with one tap from the keyboard, but for some reason, when I do that, it becomes English for me. I intend to use it for when I input in English, but there aren't many opportunities for that in my current life.

5. How to choose when you're lost—Thinking with two axes

If you're lost with voice-input LLMs, I thought it would be good to organize them using these two axes first.

Axis 1: Is text formatting necessary? Do you want to turn what you said into text as is, or do you want it structured as an "instruction"?

Axis 2: Do you prioritize reproducibility? Do you want it to come out exactly as you said, or is some correction okay?

In my case,

• Requires formatting (giving instructions on smartphone) → Typeless

• Prioritizes reproducibility (daily tasks on PC) → Aqua Voice

I settled on this way of dividing them.

It's a simple point, but having this "axis" for yourself makes it harder to get swayed when new tools come out.

6. Summary

If you choose a voice-input LLM based on "which one is the strongest," you'll usually end up lost.

Instead, in which situation you feel the least stress. Choosing based on this will result in you using it for longer.

My current answer is these two: Typeless and Aqua Voice. Superwhisper isn't bad, but the screen movement didn't suit me.

Admitting personal compatibility honestly like this makes tool selection go better. At least, that's how I feel.

#AIUtilization #VoiceInput #GenerativeAI #Practice #ForBeginners #Efficiency #LLM #HonestReview