FluidVoice, a free and open-source dictation app for macOS, has been released. It lets users switch between multiple local speech-recognition models such as Whisper and Parakeet, and processes everything locally without ever uploading audio to the cloud.
Local AI Dictation · Apple Silicon
FluidVoice: a free, on-device dictation app that never touches the cloud
Open-source (GPLv3) and permanently free, FluidVoice runs Whisper, Parakeet and Nemotron speech models entirely on the Mac — pitching itself as a local alternative to paid, subscription-based dictation tools.
3,380×
Peak real-time factor with Parakeet
50,000+
Downloads to date
$0
Cost — free & open source
Pick a model — size vs. language reach
Download size (MB) shown to scale. Language coverage noted below each column.
~250MB
Parakeet Flash
English
~500MB
Parakeet TDT v3
25 langs
~600MB
Nemotron 3.5
~40 langs
2.9GB
Whisper (Large)
up to 99 langs
Also included: Apple Speech (macOS built-in, no download) and smaller Whisper builds from 75MB.
What users like
"As good as or better than paid tools — and it's free and local"
Reported faster than rival transcribers
Solid French and German transcription
Many switching from paid subscription apps
Current limits
Intel Macs rely on Whisper — no full Parakeet speed
Fewer advanced meeting / enterprise features
iOS and Windows versions still planned
Early setup snags addressed in later updates
On-device pipeline
Voice inputglobal hotkey
→
Local modelon device
→
Fluid IntelligenceAI formatting
→
Text outputany app
All audio stays on the machine. Requires macOS 15.0 (Sequoia) or later, plus microphone and accessibility permissions.
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…