Most people speak around 130 to 150 words a minute and type around 40. If you write emails, tickets, docs or long ChatGPT prompts all day, that gap adds up to hours.
I spent 17 years in IT support. I tried the dictation apps. Most wanted an install, admin rights, a subscription before I'd tried anything, or all three. On a locked-down work laptop, that's a non-starter.
So I built Dictaro: speech-to-text that runs in a browser tab. Open the page, click the mic, talk. No install, no account, no card to try it.
How it works
- Open dictaroai.com in Chrome, Edge or Safari.
- Click record and start talking naturally.
- Stop. Your text is cleaned up, punctuated and (optionally) copied to your clipboard.
- Send it to Gmail, Outlook, ChatGPT or Claude with one click, or paste it anywhere.
It writes the way you meant it
- Spoken punctuation. Say "comma," "question mark," "open quote" or "new paragraph." Pauses add periods, and questions get question marks automatically.
- "Scratch that." Say it and the last phrase disappears.
- Filler removal. "Um," "uh" and repeated words are dropped.
- Custom vocabulary. Teach it your names and jargon once. Map "chat gpt" to "ChatGPT," "kube cuttle" to "kubectl," whatever you say a lot.
- History and export. Every dictation is saved in your browser and exports as .txt or .md.
AI Rewrite: say it rough, send it polished
Ramble out loud, then press one button:
| Mode | What it does |
|---|---|
| Clean up | Fixes grammar and flow, keeps your words |
| Polish | Clear, confident writing |
| Professional email | Notes into a ready-to-send email |
| Concise | Cuts it to the point |
| Bullet points | A ramble into a scannable list |
| Casual message | Friendly tone for Slack or Teams |
| AI prompt | Structures your idea into a clear prompt for ChatGPT or Claude |
The AI prompt mode is my favorite: talk through a bug or a feature for 30 seconds, and get a well-organized prompt instead of a vague one-liner.
How it handles your data
Text and history stay in your browser (local storage). Dictaro never receives audio; the browser's built-in speech service does the recognition (Google in Chrome, Microsoft in Edge, Apple in Safari). Text only goes to the Dictaro server when you press an AI Rewrite button, and it isn't stored.
Works on
Chrome or Edge on Windows, Mac and Linux, Safari on Mac, and Chrome on Android. iPhone is limited. 32 languages.
What it doesn't do (yet)
It doesn't type directly into desktop apps like your IDE, Word or Slack. You copy or send the text instead. It needs a connection, doesn't work in Firefox (no Web Speech API recognition there), and doesn't transcribe audio files. If you need those, an installed app like Wispr Flow or Superwhisper is the better fit.
Pricing
- Free trial: 30 minutes of dictation, every feature, no account or card.
- Pro: $7.59/month, or $73/year.
- Lifetime: $234 once.
- Teams: $6.09/user/month (3-user minimum).
Students and people with disabilities get 25% off: email robbyjsocials@gmail.com.
Try it
Open dictaroai.com, click the mic and dictate your next message or prompt. Feedback welcome, especially on which rewrite modes you'd want added.
Robby Jay is a U.S. Navy veteran and IT professional with 17+ years of help desk and IT management experience.
Top comments (1)
The browser-first approach is what caught my attention. Removing installation, account creation, and admin-rights requirements eliminates several barriers that often prevent people from even trying productivity tools.
One thing I'd be particularly interested in is the privacy boundary. Since speech recognition relies on the browser's built-in service, audio may still be processed by the browser provider rather than entirely on the device. That's an important distinction worth making very visible to users, especially those dictating confidential work information.
I also like the custom vocabulary and the AI prompt mode. Those solve two different problems: accurately capturing what you say and turning rough thoughts into something useful. The latter could be especially valuable for developers who think through problems faster than they can type them.
Curious whether you have considered an offline or genuinely on-device recognition option for users who need stricter privacy. That could make the tool useful in some environments where cloud-based dictation simply isn't an option.