README file from
GithubLive Scribe
Real-time meeting transcription in an Obsidian sidebar, powered by ElevenLabs Scribe v2 Realtime. Press start, speak (or hold a meeting), watch the transcript stream into the sidebar, and drop the parts you want into your active note. The full session is auto-saved so nothing is ever lost.
Live Scribe is an independent community plugin. It is not affiliated with or endorsed by ElevenLabs. "ElevenLabs" and "Scribe" are trademarks of their respective owners.
Features
- 🎙️ Real-time transcription from your microphone via ElevenLabs Scribe v2 Realtime (~150 ms latency, 90+ languages).
- 🌐 Auto language detection or pin a specific language (Korean, English, Japanese, Chinese, and more) for higher accuracy.
- 📝 Buffer + insert — the transcript accumulates in the sidebar; click 노트에 삽입 (Insert to note) to drop everything new at your cursor, or click a single segment to insert just that. Inserted segments dim so you always know what's left.
- 💾 Session auto-save — the complete transcript is appended live to a file in a folder you choose, so a crash or an accidental close never loses your recording.
- ⏱️ Optional timestamps — prefix each line with the wall-clock time (
14:32). - 🧩 Configurable line prefix — bullet (
-), paragraph, or quote (>). - 🔒 Client-side only — your own API key fetches a single-use token directly; no third-party server is involved. A zero-retention mode is available.
Setup
- Get an ElevenLabs API key from your ElevenLabs account (Speech-to-Text / Scribe must be available on your plan).
- Install Live Scribe, enable it, and open Settings → Live Scribe.
- Paste your API key. Optionally pick a default language, save folder, and formatting.
- Click the microphone ribbon icon (or run the command Live Scribe: 사이드바 열기) to open the sidebar.
- Open a note, click 전사 시작 (Start), grant microphone permission, and speak.
Microphone only (and how to capture a video call)
Live Scribe v1 transcribes your microphone. For in-person meetings, lectures, and dictation this works out of the box.
For a video call (Zoom / Teams / Meet), the other participants' voices come out of your speakers, not your mic. Two ways to capture them:
- Simplest: use your computer's speakers (not headphones) so your microphone picks up the room, including the remote voices.
- Cleanest: route system audio into a virtual input device — e.g. BlackHole (macOS) or VB-CABLE (Windows) — and select it as your input.
Native system/tab audio capture is planned for a future version.
Privacy & your API key
- Your API key is stored only in this vault's plugin config (
.obsidian/plugins/live-scribe/data.json). Never commit or share your.obsidianfolder — it contains your key. - Audio is streamed to ElevenLabs for transcription. Turn off 서버 로깅 허용 (Allow server logging) in settings to request zero-retention mode.
Settings reference
| Setting | Default | What it does |
|---|---|---|
| ElevenLabs API 키 | — | Your API key (stored locally). |
| 기본 전사 언어 | 자동 감지 | Auto-detect, or pin a language. |
| 세그먼트 구분 방식 | VAD 자동 | VAD auto-commits on silence (hands-free) vs manual. |
| 세션 전체 자동 저장 | ON | Append the full transcript to a session file. |
| 저장 폴더 | Transcripts | Folder for session files (auto-created). |
| 줄 접두사 | - |
Bullet / paragraph / quote. |
| 타임스탬프 표시 | OFF | Prefix each line with HH:MM. |
| 배경음 필터 | OFF | Filter background noise. |
| 서버 로깅 허용 | ON | Off = zero-retention mode. |
| 토큰 엔드포인트 (선택) | — | Advanced: use your own token proxy instead of the key. |
Platform support
Desktop is the primary target. The plugin also loads on mobile, but real-time microphone capture depends on the device supporting getUserMedia + AudioWorklet; where it isn't available, Live Scribe shows a clear message instead of failing.
License
MIT © 2026 coffee