NeuroVox Obsidian plugin
NeuroVox: an Obsidian plugin for voice notes
NeuroVox is a plugin you install inside Obsidian, the note-taking app. Record an idea while it is fresh, transcribe it into your note, and optionally process the text with AI.
The problem
A useful thought may arrive faster than you can type it. Keeping only a separate audio recording means you still have to find and transcribe the part you need later. Inside an open Obsidian note, you can capture the thought as text where the rest of your work already lives.
The solution
We built NeuroVox to capture a thought while you are in an Obsidian note, transcribe it, and put the text at your cursor. Optional AI post-processing can shape that transcript into more useful note text. You can also keep the plain transcript and edit it yourself.
How it works
-
Open the note
Choose the Markdown note where you want the spoken thought to land, and place the cursor where the transcript belongs.
-
Record your thought
Start a short recording from NeuroVox's ribbon or Command palette path. The direct floating microphone path is a separate way to record when available.
-
Transcribe the speech
NeuroVox transcribes the recording through your configured service. Depending on your settings, it can then post-process the text for your note.
-
Check the note
Read the inserted text against what you said and correct mistakes. If your recording path saves audio, check the embed in the note too.
Use NeuroVox in your app
New to plugins? Read the beginner guide. Install NeuroVox from Obsidian Community plugins on desktop, add a supported transcription provider key, then record one short thought in a test note. Review the transcript at your cursor; audio retention depends on the recording path.
Install in the Obsidian desktop app from Community plugins. Mobile is also supported where available.
Set it up
Download Obsidian if you do not have it, then create or open a test vault.
Download Obsidian ↗Open Settings, turn on community plugins, then open Community plugins and Browse.
Search for NeuroVox. Choose Install, then Enable.
Open Settings > NeuroVox. Add a transcription key from OpenAI, Groq, Deepgram, or AssemblyAI, then choose a Transcription model in that provider's group. Post-processing starts on; turn it off for a plain transcript or add an OpenAI, Groq, or OpenRouter key and choose a Post-processing model for text.
API keys and billing guide →Open a test Markdown note and start a short recording from the ribbon or Command palette. Stop and check the transcript at the cursor. If you use a direct floating microphone path, also check whether it saved an audio embed.
First use
Capture and check one voice note
Open an active Markdown note in Obsidian and use NeuroVox's ribbon or Command palette recording action.
Record one short thought from your own work, then stop the recording.
What to expect
Check for transcript text at the cursor. Audio is saved and embedded by the direct floating microphone path, but not by the default modal recording path.
Your review
Correct any transcription errors before using the note. Check for an audio embed only if your recording path saves one.
API keys and provider billing
An API key is a private credential from an AI provider account. An Obsidian plugin uses it to request a cloud service. A ChatGPT or Claude app subscription does not pay for separate API use. Check the provider’s current free tier, billing, and usage limits before sending a recording.
For speech transcription, choose OpenAI, Groq, Deepgram, or AssemblyAI. Optional text post-processing uses OpenAI, Groq, or OpenRouter. OpenRouter is our preferred starting point for compatible language models, but it does not transcribe NeuroVox audio. Anthropic and Gemini models are available for text processing through OpenRouter, not as direct NeuroVox key fields.
Choose your provider, review billing if needed, create a key in its official account site, then enter it in Settings > NeuroVox. Keep the key out of notes and chats. Choose a Transcription model from the provider group with a key; optional text processing has its own Post-processing model choice.
Official key guides
- OpenAI API quickstart — transcription or optional text processing.
- Groq API quickstart — transcription or optional text processing.
- Deepgram key guide — transcription.
- AssemblyAI key guide — transcription.
- OpenRouter API keys — optional text processing only.
Billing help for OpenAI, Groq, and OpenRouter
Frequently asked questions
Does NeuroVox save the audio?
It depends on how you record. The direct floating microphone path saves and embeds the audio in the vault. The default modal path used by the ribbon and command returns transcript text and removes its temporary segments.
Do I need an API key?
Yes for a supported cloud transcription provider such as OpenAI, Deepgram, or Groq. Provider billing and data handling depend on the service you choose.
Can I get a transcript without AI rewriting?
Yes. Turn off AI post-processing in NeuroVox settings for transcript-only use. Post-processing is on by default and may need a separate language-provider configuration.
Where does the transcript appear?
In the active Markdown note at the cursor for the default modal recording flow. Review and correct it before treating it as a record of what you said.