Murmure logo

Murmure

AI-Powered Speech-to-Text

Stop typing. Start speaking.

Murmure is an AI-powered, offline speech-to-text tool designed with Privacy first in mind and powered by NVIDIA Parakeet 🦜.

View code
Writing speed +214%|Typing errors -46%
Privacy-firstAI-poweredNo internet requiredOpen‑sourceNo telemetryNo trackingAccurate even in poor conditions25 European languages
Murmure Home Interface
What's New

New in 1.7.0

Text Commands

Apply custom commands to any transcribed text.

Multiple LLM Prompts

Save and switch between prompt configurations.

Dictionary Import/Export

CSV export, import, and medical presets.

Microphone Selection

Choose your audio input device in settings.

Text Insertion Modes

Standard, Terminal, or Direct character typing.

Signed Builds

Windows and macOS builds are now officially signed.

Digit Threshold

Configure when words are converted to digits.

Log Verbosity

Configurable log levels and quick access to logs.

Features

Discover the Power of Murmure

Murmure is evolving with powerful new capabilities for developers, medical professionals, and power users.

Local LLM Intelligence
Local LLM IntelligenceOverview
FAQ

Frequently Asked Questions

We support the following 25 european languages: Bulgarian (bg), Croatian (hr), Czech (cs), Danish (da), Dutch (nl), English (en), Estonian (et), Finnish (fi), French (fr), German (de), Greek (el), Hungarian (hu), Italian (it), Latvian (lv), Lithuanian (lt), Maltese (mt), Polish (pl), Portuguese (pt), Romanian (ro), Slovak (sk), Slovenian (sl), Spanish (es), Swedish (sv), Russian (ru), Ukrainian (uk)

Murmure is designed to be lightweight and efficient, so it can run on most devices with CPU-only desktop. However, we recommend a device with at least 2GB of RAM available and 1GB of disk space. No GPU Needed

No business model, no ads, no tracking, no data collection, no compromise. We were tired of paid speech-to-text tools that sacrifice privacy for convenience so we built one that doesn't. Our only reward is to be used by people who care about their privacy.

We use Parakeet because it is a more accurate speech-to-text model and it is more optimized for low-end devices. Parakeet is as accurate than whisper-large model (1.5B parameters) but with only 0.6B parameters making him incredibly efficient for offline and performance usage.

You can suggest issues or feature requests on our GitHub repository.