Yubi: A 6MB Mac Voice Input Tool — Just Speak to Type

A Mac Voice Input App That's Just 6MB

In an era of increasingly bloated productivity tools, Yubi takes the opposite approach — the entire app weighs in at just 6MB. Its core logic is dead simple: hold down the right Option key, start talking, and your words appear right where the cursor is. Whether you're writing an email, coding, or taking notes, no window-switching or copy-pasting required.

Built by developer Carlos Lastres, the app has earned 62 upvotes on Product Hunt (ranked #20) and is categorized under productivity, AI, and menu bar apps. It's aimed at Mac users who are tired of typing and want to capture ideas faster through their voice.

Two Hotkeys, Two Use Cases

Yubi's interaction design revolves around two modifier keys — and that's what sets it apart from typical dictation tools.

Right Option Key: Voice-to-Text

Hold down the right Option key, speak, and release — your words land at the cursor position. This is the primary use case, essentially a lighter and more immediate take on macOS's built-in dictation. For anyone who thinks out loud, this "say it, see it" input method dramatically reduces the friction between thought and screen.

macOS has a built-in Dictation feature, triggered by pressing the Fn key or microphone key twice. However, it's designed to either pop up a separate dictation interface or require focus on a specific text field to work — and by default it relies on Apple's servers (offline mode requires downloading additional language packs). Yubi's edge is its "invisible" injection via a global hotkey: whether focus is on a browser address bar, a code editor, or a terminal window, text flows in directly without the app needing to support a dictation protocol. This kind of global text injection typically leverages macOS's Accessibility API, requiring the app to have accessibility permissions in order to write to another program's text cursor.

Right Command Key: Screen Assistance

When you need help with something on screen, hold down the right Command key. This means Yubi isn't just a transcription tool — it can understand screen context and offer assistance, pushing it beyond pure voice input into AI assistant territory.

Privacy Is the Core Selling Point

In a landscape where most AI apps rely on cloud processing, Yubi emphasizes that "your voice never leaves this Mac." Local processing means no audio is uploaded to any server — a genuine draw for privacy-conscious users.

For professionals who regularly handle sensitive information — lawyers, doctors, journalists, or enterprise users — local voice processing eliminates compliance risks around data leakage. This is one of Yubi's key differentiators in a crowded field of voice tools.

The technical foundation for local speech recognition typically involves offline models. The mainstream options today include Apple's own Speech framework (which supports offline recognition) and OpenAI's open-source Whisper model — the latter can run locally without an internet connection, supports multiple languages, and has been widely adopted by lightweight desktop voice tools over the past couple of years. Yubi's 6MB footprint suggests it doesn't bundle a large speech model internally, and more likely calls into macOS's system-level local Speech recognition, or downloads a language model on first use. Which approach it actually takes directly determines the accuracy ceiling and how well it handles specialized terminology or non-English languages — the most critical technical variable when evaluating tools like this.

Low Barrier to Entry — No Credit Card Required

Yubi offers its core voice input feature for free. Users only need to sign in with an email address, and no credit card is required for the free tier. This low-friction approach makes it easy for people to try the core experience before deciding whether to pay.

That said, the official documentation doesn't yet detail what features the paid tier unlocks or what it costs. With only one user comment so far, the product is clearly in its early stages, and real-world user feedback remains limited.

The Value and Limits of a Small Tool

Yubi represents a particular philosophy in Mac app design: rather than chasing feature breadth, do one thing and do it with maximum lightness and immediacy. The 6MB size, global hotkey trigger, and local processing all point to the same goal — making voice input feel effortless.

The limitations are equally clear. Recognition accuracy, support for multiple languages and specialized vocabulary, and the actual intelligence of the screen assistance feature all need to be validated in real-world use. As an early-stage product, its potential is worth watching — but whether it can become a genuine part of daily workflows depends on how it evolves.

For Mac users curious about a voice-driven workflow, Yubi's free tier and lightweight design make it well worth trying.