YggKieirra/murmure
↗ GitHubFully local, private and cross platform Speech-to-Text with LLM Post-processing
639
Stars
62
Forks
7
Watchers
29
Open Issues
Safety Rating A
No hardcoded secrets, malicious code patterns, suspicious dependencies, or prompt injection attempts were detected. The repository is a straightforward open-source Tauri/TypeScript desktop application with a clear privacy-first purpose, transparent licensing (AGPL v3), and no red flags in the visible content.
ℹAI-assisted review, not a professional security audit.
AI Analysis
Murmure is a privacy-first, fully local speech-to-text desktop application built with Tauri and TypeScript. It uses NVIDIA's Parakeet TDT 0.6B v3 neural network model to perform on-device audio transcription with no internet connection or data collection. It supports 25 European languages and optionally integrates with a local LLM (e.g., Ollama) for post-processing tasks such as grammar correction and translation.
Use Cases
- Hands-free dictation into AI coding tools (Cursor, Claude Code, etc.) or chat interfaces
- Writing notes or capturing ideas without typing
- Offline transcription in privacy-sensitive environments
- Mass deployment of speech-to-text configuration via CLI import
- Post-processing transcriptions with a local LLM for translation or grammar fixing
Tags
Project Connections
ComfyUI-Qwen-TTS
→ComfyUI-Qwen-TTS handles text-to-speech synthesis while Murmure handles speech-to-text transcription; together they form a complete bidirectional local voice pipeline with zero internet dependency.
Voicebox
→Both are privacy-first, local-first voice desktop applications. Murmure specializes in speech-to-text transcription using Parakeet TDT; Voicebox specializes in text-to-speech synthesis with five TTS engines and voice cloning. They are functional complements sharing the same offline-first design philosophy.
local-talking-llm
→local-talking-llm assembles a full voice assistant with Whisper STT, Ollama LLM, and Chatterbox TTS. Murmure can replace the Whisper STT component with its Parakeet TDT model for improved accuracy across 25 European languages with optional grammar correction post-processing.