Steno

Record or import audio, transcribe with speaker diarization, and process through LLM prompts.

Features

• Audio recording — Record directly within Obsidian with a single command or ribbon icon.
• Audio import — Import existing audio files (mp3, mp4, m4a, wav, webm, ogg, flac) for transcription. Ideal for longer recordings made with screen off.
• Speaker diarization — Transcripts identify and label different speakers automatically.
• LLM post-processing — Run diarized transcripts through configurable prompts (meeting notes, summaries, action items, or your own custom prompts).
• Processing prompt bank — Save and manage multiple named prompts. Switch between them from settings or via command palette.
• Flexible output — Append to the current note or create a new note in a configurable folder with datetime or LLM-generated titles.
• Auto-transcribe — On mobile, new audio files added to the audio folder are automatically detected and transcribed. Pairs well with the iOS Shortcut flow for screen-off recording.
• iOS Shortcut integration — Trigger recording or import via obsidian://steno URI scheme. Bind to the iPhone Action Button for quick access.

Setup

1. Install the plugin from Community Plugins.
2. Open Settings → Steno.
3. Add your API keys:
   • Transcription: AssemblyAI or Deepgram (required for diarization).
   • LLM: OpenAI (gpt-4o), Anthropic (claude-sonnet-4-20250514), or Google Gemini (gemini-2.5-flash) (required for post-processing prompts).
4. Configure your preferred output mode and processing prompt.

Usage

Recording in-app

• Command palette (Cmd/Ctrl+P): "Steno: Toggle recording", "Steno: Start recording", "Steno: Stop recording"
• Ribbon icon: Click the microphone icon to toggle recording.
• On mobile, a modal with a large Stop button appears for easy access.

Note: On iOS, the screen must stay on during in-app recording. For longer recordings, use the import flow below.

Importing audio

• Command palette: "Steno: Import audio file for transcription"
• Select any audio file in your vault. The full pipeline (transcribe → LLM → output) runs automatically.

iOS Shortcut / Action Button

For in-app recording:

1. Create an iOS Shortcut with a single "Open URL" action: obsidian://steno?action=toggle
2. Assign the Shortcut to your iPhone Action Button.

For screen-off recording with reliable diarization (recommended):

iOS applies audio processing to in-app recordings that can reduce diarization accuracy. For reliable multi-speaker detection, use a native iOS recording via Shortcuts. Your vault must use iCloud sync for this flow.

1. Open the Shortcuts app on your iPhone.
2. Create a new Shortcut with these actions in order:
   • Record Audio — uses native iOS recording (works with screen off, no audio processing)
   • Save File — set the destination to your Obsidian vault's audio folder in iCloud Drive (e.g., iCloud Drive/Obsidian/YourVault/Steno/audio/). Enable "Ask Where to Save" or set a fixed path.
   • Open URL — set the URL to obsidian://steno?action=import
3. Assign the Shortcut to your iPhone Action Button (Settings → Action Button → Shortcut).

How it works: Press the Action Button → native recording starts (screen can turn off) → press Stop when done → audio saves to your vault → Obsidian opens and Steno automatically prompts you to select the audio file for transcription.

URI scheme

URI	Action
obsidian://steno?action=start	Start recording
obsidian://steno?action=stop	Stop recording and process
obsidian://steno?action=toggle	Toggle recording
obsidian://steno?action=import	Open audio file picker
obsidian://steno?action=import&file=path/to/audio.m4a	Import specific file
obsidian://steno?action=start&prompt=Meeting%20Notes	Start with a specific prompt

Selecting a processing prompt

• Command palette: "Steno: Select processing prompt"
• Settings: Choose the active prompt from the dropdown.
• The "Raw Transcript (no LLM)" prompt skips LLM processing and outputs the diarized transcript directly.

Network usage

This plugin sends data to external services. No data is sent without user-configured API keys.

• AssemblyAI (api.assemblyai.com): Audio is uploaded for transcription with speaker diarization.
• Deepgram (api.deepgram.com): Alternative transcription provider.
• OpenAI (api.openai.com): Diarized transcript text is sent for LLM post-processing.
• Anthropic (api.anthropic.com): Alternative LLM provider.
• Google Gemini (generativelanguage.googleapis.com): Alternative LLM provider.

Audio and transcript data is sent only to the providers you configure. Refer to each provider's privacy policy for data handling details.