# Whisper plugin for Obsidian The Whisper plugin, by nikdanilov, brings speech-to-text into [[Obsidian]] using [[OpenAI]]'s [[Whisper]] transcription model. A hotkey (Alt/Cmd+Q by default) opens a recording controls window; when you stop, the audio is sent for transcription and the resulting text lands in the vault. Existing audio files can also be uploaded and transcribed the same way. The output is configurable: transcriptions can be inserted at the cursor in the active note or saved as new notes in a chosen folder, and the audio files themselves can be kept in the vault (linked from the transcription) or discarded. Pointing the plugin at a custom API endpoint allows using a local or self-hosted Whisper server instead of OpenAI's API, keeping audio on your own machine. The plugin can also run [[Large Language Models (LLMs)]] post-processing on the raw transcript — cleaning up filler words, fixing punctuation, or reformatting the text according to a prompt — so what lands in the note is closer to usable prose than raw dictation. This turns voice capture into a first-class input channel for the vault: thoughts spoken while walking or ideas dictated faster than they could be typed arrive as searchable, editable markdown. ## References - Download: https://obsidian.md/plugins?id=whisper - Source code: https://github.com/nikdanilov/whisper-obsidian-plugin