Transcribe audio to Markdown for Obsidian

Why transcripts belong in your notes, not in another app

Most transcription tools keep your transcripts inside their own web app. You can read them there, but you cannot link them to project notes, search them together with the rest of your knowledge base, or keep them after you cancel a subscription.

Obsidian works on plain Markdown files stored on your own disk. Exporting transcripts as .md files means they live next to your other notes, work with your existing folders and plugins, and stay readable in any text editor. It suits researchers, consultants, students and product teams who already work in linked notes.

What the exported Markdown file looks like

Noteo's Markdown export is deliberately simple, so it drops into a vault without cleanup:

• YAML frontmatter with the title, the creation date when available, and the mode (local or cloud). Obsidian shows these as note properties.
• A "# Transcript" heading followed by the transcript.
• In Cloud mode, a "##" heading each time the speaker changes, using the speaker label.
• Each paragraph as a list item that starts with a bold HH:MM:SS timestamp.
• In Local mode, paragraphs cover roughly 30 seconds of speech, without speaker headings.

Worked example: a weekly meeting as a linked note in your vault

You record your weekly product sync and transcribe it in Cloud mode, so every speaker turn is labelled. Choose "Download .md" and move the file into the Meetings/ folder of your vault. Obsidian picks up the title, created_at and mode fields from the frontmatter.

Now make it yours. Add a tags property such as meeting, link the note to [[Project Atlas]] and to the people involved, and write your action items above the transcript. If you generated an AI summary in Noteo, copy it into the top of the note. Because created_at and mode are stored as properties, plugins that query frontmatter can list every meeting note by date.

Local vs Cloud: where your transcript is created

In Local mode, the Noteo desktop app extracts the audio with ffmpeg and transcribes it with whisper.cpp on your computer. Nothing is uploaded, which pairs well with a local-first Obsidian vault. You get timestamps but no speaker labels, and AI summaries are not available.

In Cloud mode, the recording is uploaded for transcription with automatic language detection and speaker labels, and you can generate AI summaries. It uses your plan minutes. Both modes produce the same Markdown structure, and the mode field in the frontmatter records which one you used.

How this compares with other ways to get transcripts into Obsidian

Copying text out of a web transcript usually loses timestamps and structure, and converting DOCX or PDF exports to Markdown means extra cleanup. Running whisper.cpp yourself gives you text, SRT or JSON output, but you still have to write the frontmatter and formatting yourself.

Noteo produces a vault-ready .md file in one click, and the same transcript can also be exported as SRT, VTT, TXT or JSON.

Limitations

The Markdown export contains the transcript only; AI summaries are not included in the file, so paste them in yourself. Noteo does not create wikilinks or tags for you, and it does not write directly into your vault folder: you download the file and move it. Local mode requires ffmpeg, a whisper.cpp binary and a model file set up in the desktop app settings.