Deepgram Transcriber for Obsidian

Transcribe audio files directly inside your Obsidian vault using Deepgram's Nova AI speech-to-text API. Right-click any audio file, get a clean markdown transcription. No external apps needed.

Why Deepgram?

Most transcription plugins use OpenAI's Whisper API. This one uses Deepgram because:

• Nova 3 is more accurate than Whisper for most use cases
• Faster processing with lower latency
• Better speaker diarization for calls and meetings
• Smart formatting handles dates, numbers, currency, and punctuation automatically
• Paragraph detection outputs readable text instead of a wall of words
• Pay-per-use pricing starting at $0.0043/min (Nova 3) with a generous free tier

Features

Single File Transcription Right-click any audio file in the file explorer and select "Transcribe with Deepgram." A companion .md note is created next to the audio file with the full transcription, duration, and model info.

Batch Transcription Run "Transcribe all audio files in current folder" from the command palette to process an entire folder of recordings at once. Already-transcribed files are automatically skipped.

Smart Output Handling

• If a companion .md note already exists for the audio file, the transcription is appended to it
• If no note exists, a new one is created with metadata and a link back to the source audio
• Output can go next to the audio file or to a custom folder

Configurable

• Choose your Deepgram model (Nova 3, Nova 2, Nova, Enhanced, Base)
• 14 language options including multi-language auto-detection
• Toggle smart formatting, paragraphs, punctuation, and speaker diarization
• Set a custom output folder for all transcriptions

Supported Audio Formats

.m4a .mp3 .wav .caf .ogg .webm .flac .aac

Installation

From Obsidian Community Plugins (Recommended)

1. Open Settings in Obsidian
2. Go to Community Plugins and click Browse
3. Search for "Deepgram Transcriber"
4. Click Install, then Enable

Manual Installation

1. Download main.js, manifest.json, and styles.css from the latest release
2. Create a folder called deepgram-transcriber in your vault's .obsidian/plugins/ directory
3. Copy the three files into that folder
4. Enable the plugin in Settings > Community Plugins

Setup

1. Go to console.deepgram.com and create a free account
2. Generate an API key (Deepgram offers $200 in free credits to start)
3. In Obsidian, go to Settings > Deepgram Transcriber
4. Paste your API key
5. Choose your preferred model and options

Usage

Transcribe a Single File

Option A: Right-click any audio file in the file explorer and select "Transcribe with Deepgram"

Option B: Open the command palette (Cmd/Ctrl + P) and run "Deepgram Transcriber: Transcribe audio file"

Batch Transcribe a Folder

1. Navigate to any file in the folder containing audio files
2. Open the command palette (Cmd/Ctrl + P)
3. Run "Deepgram Transcriber: Transcribe all audio files in current folder"
4. All untranscribed audio files will be processed. Already-transcribed files are skipped.

Output Format

Each transcription creates a markdown note like this:

# Meeting with Client

**Source:** [[Meeting with Client.m4a]]
**Transcribed:** 2026-04-11
**Duration:** 23m 45s
**Model:** nova-3

---

## Transcription

Hey, thanks for jumping on the call today. I wanted to walk through
the project timeline and make sure we're aligned on deliverables...

Settings

Setting	Description	Default
API Key	Your Deepgram API key	Required
Model	Speech recognition model	Nova 3
Language	Audio language	English
Smart Format	Format dates, numbers, currency	On
Paragraphs	Split into paragraphs	On
Punctuation	Add punctuation	On
Speaker Diarization	Label different speakers	Off
Output Location	Where to save notes	Same folder

Tips

• Turn on Speaker Diarization for calls and meetings with multiple people
• Use batch transcription to process entire folders of call recordings at once
• Set a custom output folder like Transcriptions/ to keep your vault organized
• Nova 2 is a good balance of accuracy and cost if you're processing a lot of audio

Privacy

Your audio is sent to Deepgram's API for processing. Deepgram does not store your audio data after processing by default. Review Deepgram's privacy policy for details.

Your API key is stored locally in your Obsidian vault's plugin data and is never sent anywhere except to Deepgram's API for authentication.

Building from Source

git clone https://github.com/bigfootmarketingsolutions/obsidian-deepgram-transcriber.git
cd obsidian-deepgram-transcriber
npm install
npm run build

Copy main.js, manifest.json, and styles.css to your vault's .obsidian/plugins/deepgram-transcriber/ folder.

Contributing

Issues and pull requests are welcome. If you find a bug or want a feature, open an issue.

License

MIT

Credits

Built by JJ Steelman / Bigfoot Marketing Solutions

Powered by Deepgram