Zumbo

Open source local AI voice-to-text for Mac for private STT

Mac

Productivity

Open Source

GitHub

Visit WebsiteSee on Product Hunt

Hey Product Hunt 👋 I'm Claudiu, and I built Zumbo. Most voice-to-text apps are a monthly subscription or send your voice to a server. Zumbo uses local AI models on your Mac to instantly turn what you say into text, with near perfect accuracy and complete privacy, in every app. We took over the notch, the space at the top of the MacBook screen that Apple left unused. That is where Zumbo lives: it shows it is listening while you talk, then your text, your history and your settings, without a window getting in your way. - Press a key, talk, and the text lands where your cursor is. About a fifth of a second after you stop talking. - The local AI models ship inside the app. No account, no server for your voice, and it works with the Wi-Fi off. The internet is only used to activate your license and check for updates. - 21 vocabulary packs, from developer tools to legal, medical and design. On my recordings the local AI model alone got about half of the technical terms right; with the packs, about 9 in 10. - Teach: correct a word once and it is spelled right from then on. - It cleans up as it goes: drops um and uh, trims false starts, and "two hundred" becomes "200". - Notes with reminders, and meeting mode with speaker labels. It is $15 once, with a 3-day trial, everything unlocked and no card. The whole app is open source under GPL-3 on GitHub: build it yourself, or buy the signed build with the local AI models inside. It needs an Apple silicon Mac on macOS 14 or later, and it speaks 25 European languages, English first. Three things I would love to hear: 1. Which words does it get wrong for you? Tell me and I will add them to the packs. 2. What do you use voice-to-text for most: code, email, notes or meetings? 3. Is there anything in the notch that gets in your way? Happy to get into how the local AI models run on the Mac, or the Swift side.