How to Add Voice Dictation to Obsidian (and Why It Is Not Built In)

Quick answer: Obsidian does not transcribe speech and has no built-in AI. Its audio recorder only saves an audio file, mobile dictation is just your phone keyboard's mic, and every voice or AI feature comes from a community plugin. The popular Whisper plugin works but is cloud by default and needs your own API key and per-minute fees. A simpler path is a system-wide Mac dictation tool that transcribes on your machine and drops a plain Markdown note straight into your vault folder, which fits Obsidian's own you-own-your-files ethos. The honest limit: that helps desktop capture, not phone-in-pocket capture.

People choose Obsidian for a simple, durable reason: your notes are plain Markdown files in a folder on your own machine. As Obsidian puts it, you own and control your data, and it lives on your hard disk in an open format that prevents lock-in. Even if Obsidian vanished tomorrow, your notes would still open in any text editor. That is the whole appeal.

But there is one thing Obsidian does not do, and new users are often surprised by it: it does not turn your voice into notes, and it has no AI of its own. If you want to talk your ideas in, or have them cleaned up automatically, you have to add that yourself. Here is the honest state of things, and the least painful way to fix it.

Does Obsidian have voice-to-text? Not really

Obsidian ships with an audio recorder, and that trips people up. It sounds like dictation, but it is not. The core recorder only captures an audio clip and embeds the file in your note. Nothing gets transcribed. You end up with a playable recording, not searchable text.

On your phone it feels closer, because you can tap the microphone on the keyboard and talk. But that is your phone's dictation, not Obsidian's. It is the same iOS or Android keyboard mic you get in any text field, with the same limits.

The gap is well known. One of the longest-standing requests on the Obsidian forum is simply to auto-transcribe attached audio, and an entire ecosystem of community plugins exists for no reason other than to add the speech-to-text that is not built in.

And it has no built-in AI either

The same is true for AI. Obsidian's own changelog shows no native AI and no semantic search. The popular AI features you have seen in screenshots all come from community plugins: Smart Connections surfaces related notes, and Copilot adds chat and vault search. Both are made by independent developers, not by Obsidian, and Copilot is explicit that it is not affiliated with Obsidian.

That is not a criticism. Obsidian's philosophy is to be a fast, local Markdown editor and let plugins do the rest. It just means the voice and AI parts are on you to assemble.

The plugin route, and its friction

So people reach for plugins, and they work, but each one adds setup.

The most-installed dictation plugin is Whisper by Nik Danilov, with tens of thousands of installs. It is good software. It is also cloud by default: it uploads your recorded audio to a Whisper API, it asks you to paste in your own API key, and it bills roughly half a cent per minute through OpenAI. So your audio leaves your machine, and it does not run offline unless you go out of your way to point it at a local server.

There are on-device alternatives, like Local Dictation, which transcribes on your computer through a native helper so audio is never uploaded. But it is still one more community plugin to install and configure, with a model to download first. And the AI side has the same shape: Copilot's free tier wants you to bring your own provider key, and the built-in-model version is a paid subscription.

None of this is a dealbreaker. It is just friction, stacked on the one thing you actually wanted to do: get a thought out of your head and into your vault quickly.

Why capture speed is the whole point

That friction matters more than it looks, because fast capture is the weak point of any note system, not a nice-to-have.

Tiago Forte, who popularized the "second brain" idea, describes the problem as information exhaustion: so much comes at us each day that we are left constantly anxious we are forgetting something. A second brain is supposed to relieve that. But it only works if capturing a thought is nearly effortless. Every extra step between having an idea and saving it is a chance to lose the idea. The note you never capture is the one that is gone for good.

That is exactly where a slow, multi-step voice setup quietly costs you. If dictating a quick thought means opening a plugin, waiting on a cloud round-trip, and watching your API meter tick, you will just stop doing it.

A simpler path: dictate straight into your vault

Here is the approach that keeps Obsidian exactly as it is and removes the capture friction: a system-wide dictation tool that writes plain Markdown into a folder you choose. Point it at your vault, and the notes you speak show up in Obsidian on their own.

That is what Rubber Duck does on the Mac. You hold a key and talk in any app, and it transcribes on your machine, so the audio never leaves your computer. There is no API key to paste and no per-minute fee. It files what you said as a clean, titled Markdown note in the folder you pick, which can be your Obsidian vault. Optional AI cleanup can tidy the transcript or add a summary, your choice, not a required subscription.

The nice part is how well it fits Obsidian's own values. Obsidian's whole pitch is plain-text files you own, on your disk, in an open format. A dictation tool that simply drops more plain-Markdown files into that same folder is working with that ethos, not against it. Your vault stays yours; you have just added a fast front door for your voice.

The honest limits

A few things this does not do, said plainly.

It is a Mac desktop tool, so it solves capture at your computer, not phone-in-pocket capture on the go. On your phone you are still using the OS keyboard mic, the same as before. Transcribing on-device means a one-time model download up front, in exchange for no cloud cost and no audio leaving your machine afterward. And it is not an Obsidian plugin living inside the app; it is a system-wide tool that happens to write where Obsidian reads. For some people, dictating directly into the editor with a plugin is what they want, and that is a fair reason to choose the plugin route instead.

None of that changes the core point. Obsidian is a great place to keep your thinking, and it was never trying to be a dictation app. If the missing piece for you is getting your voice into your vault without a pile of setup, you can add just that piece, on your own machine, and keep everything you already like about Obsidian.

If you want the deeper case for on-device dictation with no API bills, we wrote about that in On-device dictation with no API calls. And if you are curious why talking to AI tends to beat typing at it, see Is talking to AI faster than typing?.