Contents
Lately I've been giving instructions to AI by voice more and more.
The app I use is Superwhisper. To be precise, after some hesitation, I bought the lifetime version. 249.99 dollars, which depending on the exchange rate lands near 40,000 yen. For a voice-input app, that is straightforwardly expensive.
When I pressed buy, the feeling was: fine, just buy it.
Once bought, though, I use it without worrying about a monthly bill. I give AI long instructions constantly, so it gets a lot of use now. This post started as spoken words too.
Voice input and "transcription" are slightly different things
When people hear "turning speech into text", they often picture transcription: converting a meeting recording into a document. This is a bit different.
This is a tool for speaking instead of typing, putting text into the input field in front of you.
- Instructions to ChatGPT, Codex, Claude
- Email drafts
- Replies in Slack and other chat
- Blog post ideas
- Notes and documents
Put your cursor where you want text, speak, and your voice becomes text.
Why an "AI voice input" and not just the ordinary kind
Mac and Windows both come with free dictation. A dedicated AI voice-input app is still useful, because it does more than convert sound to characters.
When people speak, they produce "um", "no wait", "the... what was it called". AI voice input removes a fair amount of that and shapes the result into readable text.
For example, if I say:
Um, about the Cloudflare post, could you rewrite it for beginners, actually for people even earlier than beginners, I didn't know what a CDN was either so explain that bit too
it tidies away the noise and hands over something an AI can work with. Rather than composing each word, it feels like talking a problem over with the AI out loud.
Superwhisper, the one I use
My current main is Superwhisper. It runs on Mac, Windows, iPhone and iPad, and works in the input fields of many apps.
Pricing as of August 2026:
- Monthly: 8.49 dollars
- Yearly: 84.99 dollars
- Lifetime: 249.99 dollars
The lifetime licence is about 29 months of the monthly plan. Use it for more than three years and it's cheaper. Check the official pricing page for the current numbers.
What I like about it:
- Works across lots of apps (not only on Mac; on iPhone too)
- I can speak a long instruction in one go
- It shapes spoken language into something reasonably readable
- You can build modes per use: "make it email-like", "keep it exactly as spoken", "organise it as an instruction to an AI"
- Choose local processing and it works offline
Why "lifetime means never worry again" isn't quite true
Superwhisper's "Lifetime" refers to the lifetime of the product or service, not a guarantee of provision forever. Features that use cloud AI also come with fair-use conditions and limits.
So it isn't "pay once, use unlimited amounts unconditionally forever". Even so, for someone who doesn't want to add another monthly payment, the existence of a buy-once option is a real attraction. That's why I chose it.
Aqua Voice, which I used before that
Before Superwhisper I used Aqua Voice for a while. It was very good. Accurate recognition, natural-sounding text from ordinary speech, and fast.
Main pricing as of August 2026:
- Free trial: 1,000 words total
- Pro: 8 dollars a month, billed yearly
- Max: 24 dollars a month, billed yearly
(Details on the official pricing page.)
If 250 dollars up front is too much, or you want to check whether this suits you first, starting with Aqua Voice is entirely reasonable. I moved to Superwhisper because I wanted the buy-once option, not because the accuracy disappointed me.
For free, there's Genspark's Speakly
If you want to try AI voice input without spending anything, Genspark offers Speakly. The name isn't "Genspark", which makes it slightly awkward to search for.
- Free plan up to 4,000 words a week (resets weekly)
- Runs on Mac, Windows, iPhone and Android
- Removes filler words and tidies punctuation and sentences
For heavy daily use it may not be enough, but for answering "is AI voice input actually convenient?", it's very easy to try.
Codex and ChatGPT have voice input too
The ChatGPT desktop app has voice input; the shortcut is Ctrl+Shift+D (same on Mac and Windows), per the official shortcut list. There's also ChatGPT Voice for talking with the AI conversationally.
There's a big difference between that and a dedicated app, though. Voice input inside ChatGPT is fundamentally for instructing ChatGPT. Superwhisper, Aqua Voice and Speakly work in any input field on your computer: email, Slack, notes, the browser.
If you only ever talk to Codex, the built-in feature is enough. When you want it everywhere, move to a dedicated app. That order is fine.
Completely free: what Mac and Windows already have
If you'd rather not install anything, your OS already has dictation.
- Mac: enable dictation under Keyboard in System Settings (Apple's instructions)
- Windows 11: put the cursor in a field and press Windows+H (Microsoft's guide)
Both free. Try the built-in one, and move to an AI voice-input app when it starts to annoy you. That's a perfectly good starting point.
A few mistakes don't matter, because the AI fills them in
You might worry about typos, or not speaking cleanly. For instructions to an AI, you really don't need to be precious about it.
If the voice input produces:
please fix the cloudflare post so beginers can under stand it
the AI will still work out that you want the Cloudflare post rewritten for beginners. It infers intent from context.
Names, product names, amounts and dates are worth checking, obviously. But for talking something through or generating ideas, perfect input isn't required. That's exactly why voice input and AI go so well together.
Comparing them
As of August 2026, here's the summary.
| Option | Rough price | Where it works | Who it suits |
|---|---|---|---|
| Superwhisper | 8.49 dollars a month / 249.99 dollars once | Apps across Mac, Windows, iPhone, iPad | People planning to use it for years who want to buy once |
| Aqua Voice | 1,000-word trial / Pro at 8 dollars a month | Mac, Windows and others | People who want high accuracy without a big upfront cost |
| Genspark Speakly | Free up to 4,000 words a week | Mac, Windows, iPhone, Android | People trying AI voice input for the first time |
| Built into ChatGPT | Depends on your plan | Inside the ChatGPT desktop app | People who only ever talk to ChatGPT and Codex |
| Mac / Windows built-in | Free | Any input field in the OS | People who don't want to install anything |
Prices, free allowances and platform support change. Check the official sites before you subscribe.
So where should you start?
If it were me, in this order.
- Just want to try it free once → your OS dictation, or Speakly
- Want to feel "the AI shapes what I said" → Speakly or Aqua Voice
- Want daily use without a big upfront cost → Aqua Voice
- Planning to use it for years and don't want another monthly bill → Superwhisper lifetime
- Talk almost exclusively to ChatGPT and Codex → the built-in voice input
You don't have to buy a paid app first. Start with free dictation and talk to an AI: "here's the blog I want to make", "write me a post along these lines". Stumble over your words; the AI will still get it.
You no longer need to compose the perfect instruction on a keyboard before facing an AI. Just start talking. That alone changes the relationship a bit.
Prices, free allowances and platform support were checked against each service's official information on 31 August 2026. Service terms may change.
