Meeting Transcript Converter

Meeting Transcript Converter

Convert .txt and .vtt transcript files into clean Markdown notes, entirely offline. Built for meeting transcripts (Zoom recordings and Microsoft Teams exports) but works with any plain-text or WebVTT transcript.

This plugin is desktop-only and does not send data anywhere — all conversion happens locally in your vault.

Features

• Manual conversion — run "Convert transcript file (txt/vtt) to Markdown" from the Command Palette on the active .txt or .vtt file.
• Auto-conversion — optionally watch a folder and convert new transcript files as soon as they're created, with no manual step.
• VTT timestamp rendering — WebVTT cues are rendered as a bulleted list, each optionally prefixed with a formatted timestamp computed from the file's creation time.
• Microsoft Teams transcripts — Teams exports are detected automatically and converted speaker by speaker, with cue identifiers and voice tags removed.
• Frontmatter generation — output notes get YAML frontmatter with meeting_name, date, and, when detectable, duration and participants.
• Optional cleanup — delete the original transcript file automatically once it's been converted.

Usage

Manual conversion

1. Open a .txt or .vtt file in Obsidian.
2. Open the Command Palette and run Convert transcript file (txt/vtt) to Markdown.
3. The converted note is created (or updated, if it already exists) in your configured output folder.

Auto-conversion

1. Open Settings → Meeting Transcript Converter.
2. Enable Auto-convert new transcripts.
3. Set Watch folder to the vault folder you want monitored (leave empty to watch the whole vault).
4. Any .txt or .vtt file created in that folder is converted automatically. Enable Delete original file after convert if you don't want to keep the source file around.

How conversion works

TXT files are cleaned by trimming each line and removing empty lines; [Speaker] HH:mm:ss header lines also get the meeting's date, e.g. [Alice] 2026-04-05 14:32:01.

VTT files are parsed as WebVTT: the header, cue identifiers, and empty lines are skipped, and each cue becomes a bullet point. If a time format is set, each bullet is prefixed with a timestamp computed as the file's creation time plus the cue's offset, e.g.:

- **[2026-04-05 14:32:01]** Hello, this is the first line
- **[2026-04-05 14:32:06]** Second cue block here

With the time format left empty, timestamps are omitted:

- Hello, this is the first line
- Second cue block here

Every converted note starts with a title (# {name}) and YAML frontmatter containing the meeting name and date, plus duration and participant list when they can be detected from the transcript content.

Microsoft Teams transcripts are recognised automatically: they are WebVTT files whose cues name their speaker in a <v Speaker> voice tag. These convert speaker by speaker instead of as bullets, with consecutive lines from the same speaker joined into one turn:

[Kate Kan] 2026-07-31 19:31:06
Hey! One second, cannot hear you.
[Ivan Kan] 2026-07-31 19:31:11
Yeah, yeah, I was mute. Hey, hello.

Speakers found this way are also listed in the note's participants property. Detection looks only at the file's contents, so it works no matter how the export was named. Speaker lines always use YYYY-MM-DD HH:mm:ss so Teams and Zoom notes read the same; the Time format setting applies to the bulleted output above.

Settings

Installing

Manual install:

1. Download main.js, manifest.json, and styles.css from the latest release.
2. Copy them into <vault>/.obsidian/plugins/meeting-transcript-converter/.
3. Reload Obsidian and enable Meeting Transcript Converter under Settings → Community plugins.

Development

npm install         # install dependencies
npm run dev          # esbuild watch mode
npm run build         # type-check and produce a production bundle
npm run lint           # run eslint
npm run test             # run the test suite (vitest)

See the doc/ folder for detailed architecture, converter, and settings specifications.

License

This project is licensed under the GNU General Public License v3.0.