LexVoice

OBSIDIAN PLUGIN

Record, transcribe, refine, and distill— your Obsidian workspace

Record in the background, transcribe in segments or streaming, and let AI turn it into a structured Markdown summary — the raw transcript, playback timeline, and distilled knowledge all stay in your own vault.

FEATURES

From a single meeting to a knowledge base

Six capabilities around one goal: turn what was said into notes you can search, reuse, and deliver.

Meeting recording & minutes

Three audio sources: microphone only, microphone plus computer audio, or computer audio only. Segmented transcription streams in as you record; when you stop, AI merges and polishes it into structured minutes, with the master track kept intact and click-to-replay navigation.

Live outline

Generate a time-anchored topic outline while the meeting runs, updated automatically with every transcribed segment. During the meeting you can also add concepts, questions, and highlights that carry into the final write-up.

Long-audio import

Import past recordings and voice-recorder files in one click. Long audio is chunked and transcribed automatically; failed chunks enter a retry queue and can be recovered from the master track, and the body refreshes once they complete—no content is silently dropped.

Knowledge capture

Extract people, to-dos, study cards, and hot words from the minutes and write them back to your vault in one click, so every meeting settles into reusable, structured knowledge.

Semantic canvas

Turn a finished set of notes into a Canvas map: themes laid out in layers, with parallel modules and key details in view — useful for reviewing later, or for walking someone else through what happened.

Ask a question

Ask anything about any set of minutes and get answers grounded in the full text; write answers you like back into the minutes, adding your own thinking as you read.

WORKFLOW

Four steps, from voice to delivery

Scroll to watch one meeting become a structured note in your vault — every step's output stays with you.

01

Record

Start with the floating capsule; the plugin chunks audio in the background so the first segment shows up fast.

02

Transcribe

Batch segments or WebSocket streaming — captions as you speak, hot-word fixes, automatic retries on failure.

03

AI cleanup

Merge and polish by mode (meeting, interview, monologue, study) into a summary, body, and to-dos.

04

Notes & distill

Structured notes land in your vault and daily journal, then distill into people, tasks, and study cards.

RELIABILITY

Failures may happen; content must never vanish

The transcription pipeline is built to never silently drop content—every failure leaves a trace, and is retryable and recoverable.

Persistent retry queue

Failed transcription chunks automatically enter a queue and are marked as placeholders in the minutes, with automatic retries on startup; empty results are treated as failures, never mistaken for “no one spoke here.”

Auto-retry cap of 3 · adjustable 1–10

Master recording recovery

Losing a temporary segment is no problem: on retry, the full recording is re-sliced by time window—if the audio is there, the content can be recovered.

Master recording kept long-term · segment cache expires after 7 days

Resume from checkpoints

Transcription and AI cleanup each keep their own checkpoint, so a failure or a restart picks up where it left off — finished work is neither redone nor billed twice.

Per-stage checkpoints · resume after failure

One-click diagnostic report

Copy environment, configuration summary, and redacted logs in one click—keys, paths, and body text are all masked, so troubleshooting leaks nothing.

Diagnostic logs stay entirely local · redacted on write

BRING YOUR OWN API

Connect your own services

Transcription and LLM services are all configured by you—a single key can run transcription and summarization at once.

Privacy boundary

Audio and text are only sent to the transcription and LLM services you fill in yourself; the plugin itself collects and uploads nothing. API keys are stored only in the config file inside your vault, and diagnostic logs stay entirely local and are automatically redacted.

FAQ

Frequently asked questions

For fuller documentation (configuration, troubleshooting, known limitations), see theuser manual.

Can I use it on my phone?

Will my audio and data be uploaded?

What services do I need, and how do I configure them?

Will content be lost if transcription fails?

Is “who said it” accurate in the minutes?

INSTALL

Three ways to install

RELEASE

Manual install

From GitHub Release download the three files main.js, manifest.json, and styles.css, place them in your vault's .obsidian/plugins/lexvoice/ folder, then enable the plugin.

BRAT

Via BRAT

Add the repository Lynn-x/LexVoice in the BRAT plugin, and let BRAT handle installation and updates.

BUILT-IN

Built-in incremental updates

After installation, the plugin's built-in updater upgrades it in one click, with multiple download sources (including a jsDelivr mirror for mainland China) reachable even without a proxy.

System requirements: Obsidian 1.10.0 or later. Full functionality on desktop; on mobile, microphone recording, cloud transcription, and summary browsing are supported, while streaming transcription is desktop-only.

Record, transcribe, refine, and distill — all kept in your own vault.

SYSTEM REQUIREMENTS

Full features on desktop; on mobile, microphone recording, cloud transcription, and notes browsing.

Three ways to install