Fluid

FluidVoice is a free, open-source macOS dictation app with local speech models and optional local or cloud text enhancement.

The tool listed here as Fluid links to FluidVoice, a macOS dictation app that types speech into the text field you are using. It combines local transcription with optional post-processing, so you can dictate a message or a coding prompt without switching to a separate transcription window.

Dictate into the application you use

FluidVoice uses a hotkey for speech capture and requires microphone access plus accessibility permission to type into other apps. Its current macOS requirement is 15.0 Sequoia or later. The publisher also lists Intel support through Whisper models.

The app offers Write, Command, and Direct Dictation modes. App-specific prompts can change the tone or formatting of the resulting text. Check the final text before sending a message or running a command, especially if names, paths, or technical terms need exact spelling. We have not measured its transcription accuracy or the advertised latency.

Local enhancement is optional

Fluid Intelligence runs the Fluid-1 post-processing model on your Mac. The publisher describes an optional download of about 3.5GB for this local enhancement; FluidVoice can operate without it. Its purpose is to remove fillers and adjust capitalization or structure after transcription.

The product also permits optional OpenAI, Groq, and custom providers for AI enhancement. Whether a workflow remains on-device therefore depends on which providers you enable. The homepage's broad privacy language should be read alongside that configuration choice rather than treated as a guarantee for cloud-enabled use.

Match the speech model to your language

Language coverage depends on the selected model. The FAQ describes Parakeet Flash and Parakeet TDT v2 as English-only, with broader coverage for other models. Choose a model that supports your language instead of assuming each option shares the same capabilities.

FluidVoice is free and open source, with GPLv3 licensing for releases from February 23, 2026 onward. The site describes iPhone and Windows versions as coming soon. For the current Mac app, budget for model storage and review permissions before use; a local model still needs a suitable device and careful handling of the text it inserts.

Before you try it

Frequently asked questions

Does FluidVoice send dictation to a cloud service by default?

The publisher says audio and dictated text are processed locally by default. Optional cloud providers receive the content needed for the features you enable, so local processing does not describe every possible configuration.

Official source
Is FluidVoice available outside the Mac?

The official page offers the Mac app and Windows and iPhone beta access. On Mac, it requires macOS 15 Sequoia or later, microphone access and accessibility permission to type into other apps.

Official source
Are FluidVoice speech recognition and text cleanup the same model?

No. The publisher describes speech recognition and Fluid Intelligence cleanup as separate steps. On Mac, a Fluid-1 model can refine the transcript, and model choice affects hardware requirements and language coverage.

Official source