MumbleFlow icon

MumbleFlow

Fully local speech-to-text with sub-second offline transcription

Paid Dictation

Overview

MumbleFlow is a privacy-focused speech-to-text desktop app that runs entirely on your device. Powered by whisper.cpp with Metal/CUDA hardware acceleration, it delivers sub-second transcription without any cloud dependency. Press a hotkey to record, speak naturally, and your words are transcribed and pasted at the cursor in any application. A local LLM via llama.cpp handles punctuation and grammar cleanup. With support for 90+ languages, automatic language detection, and a one-time $5 price tag, MumbleFlow offers a compelling alternative to subscription-based dictation tools.

Pricing: $5

Minimum macOS: 14.0 (Sonoma)

Architecture: Apple Silicon, Intel

Key Features

  • Fully local processing - audio never leaves your device
  • Sub-second transcription powered by whisper.cpp
  • Metal hardware acceleration for fast on-device inference
  • Local LLM text cleanup via llama.cpp for punctuation and grammar
  • Global hotkey activation - works in any application
  • Support for 90+ languages with automatic detection
  • Built with Tauri 2.0 and Rust for lightweight performance
  • One-time $5 purchase - no subscription required
  • Cross-platform: macOS, Windows, and Linux

Tags

transcriptionvoice inputtext generation