Murmure is a speech-to-text application primarily focused on privacy and local processing.

Here is a summary of its main features:

Core Functionality and Privacy

• Speech-to-Text Application: Murmure is an open-source speech-to-text application that runs entirely on the user's machine.

• Privacy First and Local Processing: All transcription processing happens locally on your device. This means no data ever leaves your computer.

• No Internet Required: Because all computation is local, Murmure requires no network connection to function.

• Zero Telemetry: The application features zero tracking and zero analytics, ensuring your data remains private.

• Open Source: It is free and open source software, allowing users to inspect, modify, and contribute to the code.

Technology and Performance

• Powered by Parakeet: Murmure uses NVIDIA’s Parakeet TDT, which is described as a state-of-the-art, highly optimized, experimental transformer-based speech recognition model.

• On-Device Inference: The model is designed for fast, low-latency, on-device inference. It runs efficiently on consumer GPUs or CPUs, combining fast transcription with strong accuracy across multiple languages.

Usability and Use Cases

• Instant Transcription: Once launched, users simply start recording their voice, and the text appears instantly, processed directly on the computer.

• Extensive Language Support: Murmure supports 25 European languages. These include English, French, German, Spanish, Italian, Russian, Swedish, Polish, and others.

• Typical Use Cases: The tool is suitable for dictating to AI prompts (such as Cursor, ChatGPT, or Mistral), writing notes hands-free, or capturing creative ideas or dictation.

Advanced Features (Current and Experimental)

• Experimental API: A recent feature addition (Version 1.3.0) included an experimental API that allows Murmure to connect with external software (which can be enabled in Settings → System).