FluidVoice

FluidVoice is a free, open source voice-to-text tool for macOS that converts what you say into text you can send in any app. It’s built to keep dictation local, then add extra formatting and tone so the result reads like something you meant to write.

• On-device transcription: local speech models convert your microphone input into raw text on your Mac.
• Fluid-1 post-processing: a custom-trained model strips filler, fixes casing, formats structure, and adjusts tone to match the active app.
• System-wide dictation: one hotkey works in terminal, email, docs, chat, code editors, and other text fields.
• Offline support: transcription runs locally, so you don’t need Wi‑Fi to dictate.
• Multilanguage support: around 40 languages with some models, plus Whisper model options that support many more.
• Modes: switch between Write, Command, and Direct dictation contexts.

If you care about privacy, FluidVoice is designed so your voice and dictation processing stay on your Mac, with optional AI providers for the enhancement layer being user-controlled. It also supports both Apple Silicon and Intel Macs (Intel via Whisper models), and it’s intended for low-latency use so text appears quickly while you’re speaking.

FluidVoice is open source (GPLv3 from 2026-02-23 onward) and free forever, which makes it a fit if you want something you control instead of a paid cloud dictation subscription. It also uses an Apple-focused pipeline (CoreML and Metal), which aims to keep transcription responsive during both short dictation and longer sessions.