marlene
v2.0.0
Published
Speech to text Node app with Voxtral
Readme
Marlène
Speech to text Node app with Voxtral (inspired from Mistral Realtime transcription docs).
Start Marlene with your Mistral API key to transcribe the audio input into a local file. Example of usage: Combined with softwares that constantly save your changes (e.g. Obsidian), you can work on your document by mixing typing and speaking.
Requirements
Runs with Node.js >= 24 and pnpm.
Usage
pnpm dlx marlene --api-key <YOUR-MISTRAL-API-KEY> --output "path/to/the/file/to/write"Options
--api-key: Mistral config/ Your Mistral API key (required, can be read fromprocess.env)--model: Mistral config/ Model ID (default tovoxtral-mini-transcribe-realtime-2602)--base-url: Mistral config/ API base URL (default towss://api.mistral.ai, can be read fromprocess.env)--encoding: Audio config/ Audio encoding (default toAudioEncoding.PcmS16le)--sample-rate: Audio config/ Sample rate in Hz (default to16000)--output: Output config/ Filename (default totranscription.txt)
