FluidVoice

FluidVoice is a free, open-source dictation application for macOS that enables on-device voice-to-text transcription using local AI models. Designed for users seeking privacy and efficiency, it keeps all data processing on the user's Mac, ensuring that no data leaves the device. The app is compatible with both Apple Silicon and Intel Macs running macOS 15.0 or later, and is available for installation via direct download or Homebrew.

The application integrates the Fluid-1 AI model, which serves as a local intelligence layer for post-processing dictation. This model polishes rough transcriptions, adapts tone based on the active application, formats structure, corrects casing, and handles elements like dates, names, and numbers. FluidVoice supports over 40 languages and offers predefined modes such as Write Mode, Command Mode, and Direct Dictation, allowing users to switch context instantly. The system recognizes which application is in focus and can apply customized tone profiles automatically, so dictation fits the context—whether it’s a casual Slack message, a formal email, or structured notes for platforms like GitHub.

FluidVoice operates entirely offline, leveraging local speech models such as Nemotron Speech 3.5, Parakeet Flash, Parakeet TDT v3/v2, Cohere Transcribe, Apple Speech, and various sizes of Whisper. Users can also choose between local and cloud-based AI providers, including OpenAI and Groq, for post-processing if desired. The app is optimized for low latency and high speed, with transcription that is significantly faster than typing and perceived latency of under 100 milliseconds. It supports long-form transcription sessions without requiring manual chunking, and works system-wide—allowing dictation into any text field across apps like Cursor, Slack, Notion, Chrome, Discord, Ghostty, Mail, Messages, VS Code, Xcode, and Notes.

FluidVoice is licensed under GPLv3 and is free to use without tiers or restrictions. Its open-source codebase is community-accessible, making it suitable for developers, multilingual teams, and privacy-conscious users who require accurate, context-aware dictation on macOS.