FluidVoice Review: Fastest macOS Offline Dictation App

If you've ever wanted to dictate text on your Mac without sending your voice to the cloud, FluidVoice is the open-source answer. It's a blazing-fast, offline dictation app that runs entirely on your machine, supporting multiple speech models and even a local AI enhancement layer called Fluid Intelligence. With over 3,900 GitHub stars and a thriving community, it's quickly becoming the go-to voice-to-text tool for privacy-conscious macOS users.

What is FluidVoice?

FluidVoice is a free, open-source (GPLv3) dictation application built specifically for macOS. It leverages on-device speech recognition models like Nemotron Speech 3.5 and Parakeet to convert your voice to text locally. The app also offers a separate, privately maintained local AI runtime called Fluid Intelligence that adds smart formatting, context-aware capitalization, and post-processing — all without ever sending data off your machine. Whether you're a developer writing code, a student taking notes, or a professional dictating reports, FluidVoice aims to be the fastest, most private option on macOS.

Why it caught my eye

The problem FluidVoice solves is simple but painful: most dictation tools either rely on cloud processing (which raises privacy concerns and requires internet) or are sluggish and inaccurate. Apple's built-in dictation is okay but limited in language support and doesn't offer the latest AI models. If you're a writer, programmer, or anyone who types a lot, dictation can speed up your workflow — but only if it's fast, accurate, and respects your privacy. FluidVoice checks all those boxes by running entirely on your Mac. It supports a variety of models, from the fast Parakeet to the multilingual Nemotron, so you can choose what fits your needs. The optional Fluid Intelligence layer is a nice bonus for those who want AI-powered enhancements without an internet connection.

How it works

FluidVoice is built on a modular architecture that separates the core dictation engine from the UI and AI enhancement layer. Here's a quick overview of its key concepts:

• Speech Models: The app downloads and runs models like Parakeet Flash (fastest), Nemotron Speech 3.5 (multilingual), or Apple Speech (built-in). Models are stored locally and loaded on demand.
• Fluid Intelligence: A separate, closed-source runtime that post-processes transcribed text for smart formatting, capitalization, and punctuation. Runs entirely on-device.
• Command Mode: Listen for voice commands to launch apps, run shortcuts, or trigger system actions — think of it as a voice-controlled macro system.
• Write Mode: Injects transcribed text directly into any text field using macOS accessibility APIs. Works across apps.
• Global Hotkey: Start/stop dictation from anywhere with a customizable keyboard shortcut.
• Overlay: A live preview widget that shows your transcription in real time, optionally fitting around the MacBook notch.

All processing happens on your machine; no data leaves unless you opt into a cloud AI provider (like OpenAI or Groq) for enhancement.

Quick start

Getting started with FluidVoice is straightforward. Install via Homebrew:

brew install --cask fluidvoice

Alternatively, download the latest release from GitHub and drag the app to your Applications folder.

Once installed, launch FluidVoice. You'll be greeted by an onboarding flow that guides you through:

1. Granting microphone and accessibility permissions.
2. Selecting your preferred speech model (e.g., Parakeet Flash for English, Nemotron for multilingual).
3. Setting up a global hotkey (I chose Option+Space).

That's it! Press your hotkey, speak, and see text appear in any text field. The default model is fast enough for real-time dictation, and you can tweak settings from the menu bar icon.

Real-world example

Imagine you're writing an email in Apple Mail. With FluidVoice, you can compose entirely by voice. Here's how:

1. Open Apple Mail and place your cursor in the email body.
2. Press your global hotkey (e.g., Option+Space). The overlay appears, showing "Listening...".
3. Dictate your message: "Dear team, just a quick update on the project. The deadline has been extended to next Friday. Please let me know if you have any questions. Best, Alex."
4. Press the hotkey again to stop dictation. The text appears in the email body, properly formatted with punctuation thanks to the AI enhancement.

You can also use Command Mode to send the email: say "Send email" after dictating. FluidVoice can be configured to trigger macOS shortcuts or automations via its command system.

If you often write in specific contexts (e.g., code editors, email, notes), you can set up per-app configurations with different prompt sets for the AI enhancement layer. This means your dictation style adjusts automatically — concise for code comments, formal for emails.

Pros and cons

Pros

• Fully offline — no internet required for core dictation.
• Multiple state-of-the-art models support, including Parakeet and Nemotron.
• Fluid Intelligence adds smart formatting and context awareness locally.
• Command Mode enables voice control of macOS, reducing keyboard reliance.
• Active development with frequent updates (latest v1.6.0).
• Open-source core with permissive GPLv3 license.

Cons

• Fluid Intelligence is closed-source and may be removed or changed in the future.
• Requires macOS — no Windows or Linux support yet (iOS and Windows planned).
• Some models are large (e.g., Nemotron ~670MB) and need significant storage.
• Accessibility permissions can be tricky to set up for some users.

Alternatives

• macOS Built-in Dictation: Free, built into System Settings, but limited to Apple's model and requires internet for enhanced dictation. Less language support and no AI enhancement.
• MacWhisper: A popular open-source Whisper-based transcription app for macOS. It's reliable but slower than FluidVoice's Parakeet implementation and lacks Command Mode.
• Dragon NaturallySpeaking: Industry-standard dictation software, but expensive, cloud-dependent, and not fully offline. Available on Windows and macOS, but macOS version is less feature-rich.

My verdict — should you use it?

If you're on macOS and value privacy, speed, and the latest speech models, FluidVoice is a no-brainer. It's the fastest offline dictation app I've tested, and the optional Fluid Intelligence layer is a game-changer for formatting. I'd recommend it to writers, developers, and anyone who types all day — especially if you're uncomfortable with cloud-based dictation. Skip it if you need Windows support, prefer Apple's free but limited built-in dictation, or require a fully open-source stack (Fluid Intelligence is private). Otherwise, give it a try — the Homebrew install takes seconds, and the difference in speed and accuracy is remarkable.