Obsidian Stats logoObsidian Stats

Voice MD

by denizokcu
5
4
3
2
1
New Plugin
Install Code

Description

This plugin has not been manually reviewed by Obsidian staff. Record audio and transcribe it to markdown using OpenAI Audio Transcription API.

Reviews

No reviews yet.

Stats

stars
downloads
0
forks
0
days
NaN
days
NaN
days
0
total PRs
0
open PRs
0
closed PRs
0
merged PRs
0
total issues
0
open issues
0
closed issues
0
commits

Latest Version

Invalid date

Changelog

See all version on GitHub

README file from

Github

Voice MD

Record your voice, get markdown. Powered by OpenAI's latest audio models — works on desktop and mobile.

Transcribe directly into your active note, or use meeting mode to identify speakers in conversations. Optionally run GPT post-processing to turn raw transcripts into structured notes with headings, lists, and paragraphs.

GitHub release

Quick start

  1. Install from Obsidian Community Plugins (search "Voice MD")
  2. Add your OpenAI API key in Settings → Voice MD
  3. Click the microphone icon in the ribbon or run "Start voice recording" from the command palette
  4. Speak. Stop. Text appears at your cursor.

Features

Voice-to-markdown

Record audio and get a transcription inserted directly at your cursor position. Uses gpt-4o-mini-transcribe — fast, accurate, and supports automatic language detection. You can also force a specific language (en, de, fr, etc.) in settings.

Meeting mode

Enable the "Meeting Mode" checkbox in the recording modal to identify speakers automatically. Uses gpt-4o-transcribe-diarize — best with 2–6 speakers and recordings over 30 seconds.

Output looks like this:

**Speaker A:** Let's review the Q3 numbers.

**Speaker B:** Revenue was up 12%, mostly driven by the enterprise segment.

**Speaker A:** What about churn?

Post-processing (smart formatting)

Enable the "Post-Processing" checkbox in the recording modal to run GPT formatting on your transcript. This creates two files in a Voice Transcriptions/ folder:

  • transcription-2025-05-13-143022-raw.md — verbatim transcript
  • transcription-2025-05-13-143022.md — formatted version with headings, lists, and structure

The formatted file links back to the raw version so you never lose the original. You can configure the GPT model (gpt-4o-mini, gpt-4o, etc.) and write a custom formatting prompt for your use case.

Both checkboxes default to your global settings and persist changes back — so you can toggle per recording without losing your defaults.

Mobile support

Works on iOS and Android. Audio format detection adapts to your platform (WebM on desktop, MP4 on iOS, OGG/WAV as fallback).

Settings

Settings → Voice MD

Setting Description Default
OpenAI API key Required. Get one here —
Max recording duration Maximum seconds per recording 300
Auto-start recording Start recording immediately when the modal opens Off
Language Force a language code, or leave blank for auto-detect Auto
Enable post-processing Default for the post-processing checkbox Off
Chat model GPT model used for post-processing gpt-4o-mini
Custom formatting prompt Override the default formatting instructions —

Installation

Community plugins (recommended)

  1. Settings → Community plugins → Browse
  2. Search "Voice MD"
  3. Install and enable

Manual

  1. Download the latest GitHub release
  2. Copy main.js, manifest.json, and styles.css to <vault>/.obsidian/plugins/voice-md/
  3. Enable in Settings → Community plugins

Beta (BRAT)

  1. Install the BRAT plugin
  2. BRAT settings → Add Beta plugin → DenizOkcu/voice-md

Privacy

  • Audio is sent to OpenAI for transcription only — never stored on disk or written to your vault
  • Post-processing sends the transcript text to OpenAI if enabled
  • Your API key stays in local Obsidian storage
  • No telemetry, no tracking, no third-party services

Troubleshooting

Problem Fix
No API key error Add your OpenAI API key in settings
Recording won't start Grant microphone permission to Obsidian in your OS settings
Transcription fails Check that your API key is valid and your OpenAI account has credits
No speaker labels Meeting mode works best with 2–6 speakers and recordings over 30 seconds

For anything else, open an issue.

License

MIT

Sponsors