Magic Mic

Record, transcribe, and summarize audio directly into Obsidian with custom assistants.

⭐️ Run custom summary assistants

Summary assistants turn your transcribed memos into useful notes. Set up one assistant for journal entries and another for meeting notes - it's completely up to you!

📃 Transcribe with spelling and style hints

Seed Magic Mic with a sample of names and acronyms to prevent annoying misspellings. Be sure to read OpenAI's docs to get the most out of this feature.

🔴 Record as long as you like

Magic Mic handles long running audio gracefully and intelligently following OpenAI's best practices.

🎧 Transcribe and summarize outside audio

Use your summary assistants on audio that wasn't recorded within Obsidian. Just add the audio to your vault, open it, and run the 'Transcribe and summarize' command with one of your assistants.

Getting started

Configure an OpenAI API key in the Magic Mic settings. You can get one from https://platform.openai.com/api-keys.

Then, click the Magic Mic icon (the microphone) to begin recording. Click the icon again to pause or finish recording. A new note will be created with the audio file, a transcript, and summary generated by your chosen assistant. The note and audio file are stored in the vault according to your Obsidian settings.

Now that you've gotten the hang of it, try controlling Magic Mic via commands! You can start, pause, resume, cancel, or finish recording.

Settings and customization

• Save audio

  • Whether to save audio files into your vault after recording. If true, audio will be saved according to your "Default location for new attachments" setting
  • Default true

• Link audio

  • Whether to link the audio file into the new note; ignored if 'save audio' is false
  • Default true

• Open API Key

  • An OpenAI API key obtained from https://platform.openai.com/api-keys

• Speech to text hints

  • Hint the transcription with words, acronyms, or names that are likely to appear in your audio, or with stylized text you want the transcript to match. For longer transcriptions that require more than one API call, the prompt will be prepended to the final tokens of the previous response to improve consistency across segments.
  • Default ''

• Summary assistant model

  • The OpenAI model to use in summarizing the transcript
  • Default gpt-4o

• Summary assistants

  • Summary assistants turn your transcribed memos into useful notes. Provide a name and a prompt for each. Add multiple assistants for different purposes, and choose between them when you run Magic Mic.
  • Default (see below)