Getting Started

Setup Guide

Install the Phonolite plugin and start turning voice notes into structured Obsidian markdown.

Quick Start

Install the plugin

Phonolite is pending review for the official Obsidian Community Plugins directory. Install it now via BRAT (recommended) or manual download:

Installation instructions →

Create an account and get your API key

Sign up at phonolite.rocks using your Google account. Then go to your dashboard and create an API key. It will look like pk_....

Note: Free accounts include 10 minutes of transcription per month for the first 3 months. Upgrade to Pro ($5/mo) for unlimited use.

Paste your API key into plugin settings

In Obsidian, open Settings → Phonolite and paste your key into the API key field. Leave Server URL as-is unless you are self-hosting.

Choose your transcription mode and start recording

On desktop, the plugin will automatically download a local Whisper model (~40 MB) in the background for private on-device transcription. You can monitor the download under Settings → Phonolite → Model status.

On mobile, transcription always uses the cloud (no setup needed).

Once ready, click the microphone icon in the Obsidian ribbon (left sidebar) to start recording. Click it again to stop — Phonolite will transcribe, structure, and save a new note automatically.

Transcription Modes

💻

Local (on-device)

✓ Audio never leaves your device
✓ Works offline
✓ Desktop only (Windows, Mac, Linux)
✓ Choice of Tiny (~40 MB) or Base (~145 MB) model
Requires ~40–145 MB disk space
First-time model download required

☁️

Cloud (Groq Whisper)

✓ Works on mobile and desktop
✓ No setup or model download
✓ Slightly more accurate than Tiny model
Audio sent to Groq for processing
Counts against monthly usage quota
Requires internet connection

Note: On desktop, local mode is used automatically when the model is downloaded. Toggle Force cloud transcription in settings to always use the cloud instead.

Note: If local transcription fails, the plugin will ask whether to send the recording to the cloud or discard it.

Downloading the local model

The model downloads automatically when the plugin first loads on desktop. To manage it manually:

Open Settings → Phonolite
Find the Model status indicator
If not downloaded, click Download now
Choose Tiny (faster) or Base (more accurate) from the Model size dropdown — changing this will download the new model

Settings Reference

General

API keytextdefault: (empty)

Your Phonolite API key. Get one from your dashboard at phonolite.rocks. Format: pk_...

Server URLtextdefault: https://phonolite.rocks

Backend API URL. Leave as default unless you are running a self-hosted deployment.

Transcription (desktop only)

Model sizedropdowndefault: Tiny

Tiny (~40 MB) — faster transcription, slightly lower accuracy. Base (~145 MB) — more accurate, slower. Switching model size triggers a new download.

Model storage pathtext

Where the Whisper model files are cached on disk. Defaults to .obsidian/plugins/phonolite/models/. Use the Open folder button to navigate there.

Model statusread-only

Shows the current model state: ✅ Ready, ❌ Not downloaded, ⏬ Downloading, or ⚠️ Download failed. Includes a Download now or Retry button as appropriate.

Force cloud transcriptiontoggledefault: off

When enabled, always sends audio to Groq cloud transcription instead of using the local model. Useful if you prefer cloud accuracy or want to conserve disk space.

Note Generation

Output foldertextdefault: (vault root)

Vault-relative folder where generated notes are saved. Leave empty to save in the vault root. Example: Phonolite Notes

Recordings foldertextdefault: phonolite/recordings

Where raw audio recordings (.webm) are saved. Leave empty to skip saving recordings.

Transcripts foldertextdefault: phonolite/transcripts

Where raw transcripts (.md) are saved before AI structuring. Leave empty to skip saving.

Note templatetextarea

Markdown template used to render the final note. Supports these tokens: {{title}}, {{summary}}, {{sections}}, {{actionItems}}, {{tags}}, {{date}}, {{transcript}}. See the Note Template section below.

Custom prompttextareadefault: (empty)

Optional text prepended to the AI system prompt to guide note generation. Example: Always respond in Spanish. or Format action items as a numbered list.

Tools

Transcribe audio filebutton

Opens a file picker to select any audio file (.webm, .wav, .mp3, .ogg, .m4a) from your vault and runs the full transcribe → convert → save pipeline. Useful for reprocessing old recordings.

Convert transcript to notebutton

Opens a file picker to select a raw transcript (.md) from your vault and converts it into a structured note. Skips transcription — useful if you already have the text.

Commands & Shortcuts

Command	Description
Start recording	Begin capturing audio from your microphone
Stop recording	Stop recording and run the full pipeline
Transcribe audio file	Pick an audio file from your vault and process it
Convert transcript to note	Pick a raw transcript and convert it to a structured note

Note: The microphone ribbon icon (left sidebar) toggles recording on/off. You can also assign a hotkey to Start/Stop recording via Settings → Hotkeys — search "Phonolite". The default hotkey is Ctrl+Shift+R (Windows/Linux) or Cmd+Shift+R (Mac).

Note Template

The note template controls how generated notes are formatted. Edit it under Settings → Phonolite → Note template.

Available tokens

Token	Replaced with
{{title}}	AI-generated note title
{{summary}}	One-sentence summary of the recording
{{sections}}	Structured sections with ## headings and content
{{actionItems}}	Bullet list of action items extracted from the note
{{tags}}	Comma-separated tags (e.g. meeting, project-x)
{{date}}	Current date in YYYY-MM-DD format
{{transcript}}	Raw transcript text

Default template

---
tags: {{tags}}
date: {{date}}
source: phonolite
---

# {{title}}

> {{summary}}

{{sections}}

## Action Items

{{actionItems}}

Ready to try it?

Get Started View on GitHub