Fluid NEW

FluidVoice is an open-source voice-to-text dictation app for macOS that runs local speech models and an optional on-device AI post-processor (Fluid-1) for context-aware formatting and capitalization.It provides system-wide input via a hotkey and overlay, letting developers, writers, and accessibility users dictate into terminals, code editors, email, chat, and documents.

FluidVoice supports multiple speech engines (Nemotron, Parakeet, Whisper, Cohere, Apple Speech) and multilingual transcription across model-specific language sets.Fluid-1 performs local post-processing to adapt tone by app, fix casing, format dates/numbers, and structure output without sending data off-device.

The app is optimized for Apple Silicon and Intel macs with CoreML/Metal acceleration for low-latency, real-time transcription and long-form sessions.Configurable modes (Write, Command, Direct) and per-app prompt profiles let teams control output style and integration with custom AI providers.

Key Features

• On-device AI post-processing (Fluid-1) for smart formatting, context-aware rewriting, capitalization, and handling dates/names/numbers
• Local speech transcription with multiple optimized models (Nemotron, Parakeet, Cohere, Apple Speech, Whisper) and multilingual support
• System-wide input via hotkey and overlay to type dictated text into any app (email, docs, terminal, code editors)
• Predefined modes (Write Mode, Command Mode, Direct Dictation) with history and command capabilities
• Low-latency, Apple Silicon–optimized inference using CoreML/Metal for real-time transcription and long-form sessions

Use Cases

• Dictate and format polished technical documentation and code comments directly inside your editor using FluidVoice's code-editor dictation and the on-device Fluid-1 post-processor for context-aware formatting and capitalization, enabling hands-free coding, instant real-time transcription and local-first privacy
• Transcribe multilingual interviews, meetings and lectures in real time with system-wide hotkey, multi-engine support and a system-wide dictation overlay, then apply on-device post-processing to add punctuation and correct casing without sending audio to the cloud for secure, accurate transcripts
• Create well-structured emails, blog posts and notes by speaking naturally into FluidVoice, leveraging multilingual transcription and Fluid-1's contextual formatting to produce publication-ready text, switch languages on the fly and keep all data on-device for privacy

Who is it for?

• Privacy-conscious users
• Developers
• Writers
• Multilingual users
• Mac enthusiasts