Say

pending

by Jordan Egstad

Record audio, transcribe with Whisper, and summarize + tag with Claude AI.

Updated 4d agoMITDiscovered via Obsidian Unofficial Plugins
View on GitHub

Say — Voice Notes for Obsidian

Record a voice memo, get a structured note. Say uses OpenAI Whisper to transcribe your audio and Claude to turn the transcript into a clean, organized Obsidian note — with a summary, key points, tasks, and tags.

Desktop only. Requires microphone access and API keys for OpenAI and Anthropic.


How it works

  1. Click the mic icon in the ribbon (or run Record voice note from the command palette)
  2. Speak — a live waveform shows the mic is active
  3. Click Stop & Process
  4. The transcript is sent to Whisper, then to Claude
  5. The result is inserted at your cursor in the active note

Output format

Each recording produces a structured markdown block:

# Title of the note

## Summary
2–4 sentence summary of key points, decisions, and named entities.

## Key Points
- Point one
- Point two

## Tasks
- [ ] Action item extracted from speech
- [ ] Another task

## Transcript
> [!note]- Full Transcript
> The raw transcription text...

## Tags
#person/name #place/city #topic/label

Setup

1. Install the plugin

Install via Obsidian's Community Plugins browser, or manually copy main.js, manifest.json, and styles.css into .obsidian/plugins/voice-notes-ai/.

2. Add your API keys

Go to Settings → Say and enter:


Settings

SettingDescription
Anthropic API KeyUsed to call Claude for summarization
OpenAI API KeyUsed to call Whisper for transcription
ModelChoose Claude Opus (most capable), Sonnet (fast & smart), or Haiku (fastest)
Custom InstructionsThe full prompt sent to Claude — edit to change the output format entirely
Extract tagsToggle the ## Tags section on or off
Include full transcriptToggle the ## Transcript callout on or off

Custom Instructions

The Custom Instructions field is the complete system prompt sent to Claude. The default prompt produces the structured format shown above. You can replace it entirely to change the output — for example to use a different language, different sections, or a completely different structure.


Privacy

  • Audio is sent to OpenAI (Whisper) for transcription
  • The transcript is sent to Anthropic (Claude) for summarization
  • Your API keys are stored locally in your vault at .obsidian/plugins/voice-notes-ai/data.json
  • No data is sent to any other service, and nothing is logged or tracked

Requirements

  • Obsidian v0.15.0 or later
  • Desktop only (uses the Web Audio API and MediaRecorder)
  • An active OpenAI API key (for Whisper transcription)
  • An active Anthropic API key (for Claude summarization)

License

MIT

For plugin developers

Search results and similarity scores are powered by semantic analysis of your plugin's README. If your plugin isn't appearing for searches you'd expect, try updating your README to clearly describe your plugin's purpose, features, and use cases.