← Back to Blog

Short answer: Obsidian has no built-in dictation button, but a system-wide voice-to-text tool types straight into the editor pane with nothing installed in your vault. Click into a note first, speak the prose, and keep typing the syntax: wikilinks, tags and code fences are keyboard work.

Most Obsidian vaults have the same problem, and it is not organisation. It is that the notes are too short. There are hundreds of files with good titles and three lines underneath, links pointing at pages that were never written, and daily notes with a bullet that says "think about this later" and nothing about what "this" was.

That is not a discipline failure. It is an input cost. You had the whole thought in your head at the moment you created the note, and typing it out would have taken four minutes, so you typed the title and moved on. Speaking the same thought takes forty seconds. When the cost of writing the body drops below the cost of leaving a stub, the vault starts filling itself.

This guide covers how to dictate in Obsidian on macOS and iOS: which surfaces are actually worth speaking, what you should still type, and the specific things about Obsidian, Markdown syntax, Vim mode, hotkey conflicts, that trip people up on the first try.

Obsidian does not need a dictation plugin

Obsidian is a Markdown editor over a folder of plain text files. It ships no microphone button, and the community plugin ecosystem for voice input tends to mean more moving parts inside a vault you deliberately keep simple: another plugin to update, another thing that touches your files, another API key to manage.

You do not need any of that, because the editor pane is a normal text field. A system-wide dictation tool that types at your cursor works in Obsidian for exactly the same reason it works in Mail, Slack or a browser text box. It does not know Obsidian exists, and Obsidian does not need to know it exists.

That distinction matters more than it sounds. Nothing gets installed in your vault, no plugin is added to your synced .obsidian folder, and your notes stay exactly what they were: plain text files on disk that any other program can read.

Voice Keyboard Pro works this way on the Mac. It lives in your menu bar, and you hold a hotkey, speak, and release. The text appears wherever your cursor is, including inside an Obsidian note, a search box, or the rename field on a file. On iPhone it is a keyboard with a microphone built in, which means it works inside Obsidian Mobile the same way it works in Messages.

Setup: the four things to get right first

1. Click into the note before you speak

This is the single most common reason people think dictation "does not work in Obsidian." Obsidian has a lot of focusable regions: the file explorer, the search pane, the graph view, the command palette. If focus is in the sidebar, your words have nowhere to go. Click into the editor, confirm you can see a blinking cursor, then start.

2. Pick a hotkey Obsidian has not claimed

Obsidian is a keyboard-first app, and heavy users remap a lot. Before you settle on a dictation hotkey, open Settings, then Hotkeys, and search for the combination you are considering. If it is already bound to Quick Switcher or Insert Template, one of the two will win and it will not be obvious which. A key that Obsidian is unlikely to ever want, such as a function key or a right-side modifier held on its own, avoids the argument entirely.

3. If you use Vim key bindings, be in insert mode

Obsidian's Vim mode is popular, and it produces the most confusing possible failure: you speak a sentence, and instead of text you get a cascade of commands. Half your note vanishes, something gets yanked, the view jumps. Nothing is broken. Dictated text arrives as keystrokes, and in normal mode keystrokes are commands. Press i first. If you have already done the cascade version, undo repeatedly before you do anything else.

4. Do a thirty-second test in a scratch note

Make a note called scratch, put the cursor in it, and dictate two sentences. This tells you three things at once: whether the hotkey survives Obsidian's key map, whether punctuation is landing the way you expect, and how the text looks in Live Preview versus Source mode. Thirty seconds here saves you from debugging in the middle of a real note.

The five surfaces worth dictating

1. The daily note

The daily note is the highest-value dictation surface in any vault, because it is the one you are most likely to skip. Two passes work well. In the morning, speak what you intend to do and, more usefully, why, since the reasoning is the part that is never typed. In the evening, speak what actually happened, what surprised you, and what you would tell yourself tomorrow morning.

The evening pass is where dictation earns its place. Typed end-of-day notes decay into three bullets because you are tired. Spoken ones stay in full sentences, and full sentences are what make a note useful eighteen months later when the shorthand has stopped meaning anything.

2. Literature notes, in your own words

The standard advice for reading notes is to restate the idea in your own words rather than quoting. The standard reason people quote instead is that pasting is instant and restating is slow. Dictation removes that gap. Finish a chapter, put the cursor under the reference, and say what the argument was, whether you believe it, and what it connects to in the vault.

This produces the kind of note that is worth linking to, rather than a highlight dump that you will never reread.

3. The empty body under a good title

Every vault has these. A note called "why our onboarding flow leaks users at step three," created in a flash of clarity, containing nothing. Open ten of them, and speak two or three paragraphs into each. Most will take under a minute, because the thinking was already done. You are just paying the transcription cost that stopped you the first time.

This is the single fastest way to make an existing vault more valuable, and it does not require reorganising anything.

4. Meeting and call notes

Two different jobs hide here. During a meeting, you are usually typing fragments, and a keyboard is fine for that. After it, in the ten minutes where the useful version gets written, dictation is much faster: what was decided, who owns what, what was left unresolved, and what you personally need to do before Thursday.

On the Mac, Meeting Mode can capture the meeting itself with speaker detection and produce AI notes, and calendar meeting detection means it can start when your meeting does. Treat what comes out as raw material for a note, not as the note. Whether recording is appropriate is a call for you and the people in the room, not for a tool.

5. Capture on the phone

Obsidian Mobile is where the thoughts that never reach the vault get lost. Walking, driving as a passenger, standing in a queue: the idea arrives, and typing it on a phone keyboard is unpleasant enough that you promise to write it later and do not. A keyboard with a microphone turns that into a ten-second capture into today's daily note.

Voice Edit helps here too. Phone editing is the worst kind of editing, and being able to speak a correction rather than place a cursor between two characters with a thumb makes the difference between fixing a note and leaving it wrong.

Speak the sentences, type the syntax

This is the rule that makes dictation in Obsidian pleasant instead of frustrating. Markdown is text, but Markdown syntax is not prose, and speaking punctuation-dense structures aloud is slower and less reliable than typing them.

Speak: paragraphs, explanations, arguments, summaries, questions to yourself, meeting outcomes, anything that would be a sentence if you said it to a colleague.

Type:

The workflow that follows from this is straightforward: dictate the paragraph first, then go back and bracket the terms that should be links. It reads better too, because you are choosing links based on the finished sentence rather than interrupting yourself mid-thought to open a popup.

Headings and bullets

Structure sits in between. Speaking "new line" and "new paragraph" works well and keeps you in flow. Speaking bullet markers is more mixed: it is usually faster to type the dash, speak the item, press return, and let Obsidian continue the list for you. We covered the general version of this in dictating bullet points and lists, and the same pattern applies here.

The proper noun problem, and the fix

A vault is dense with words no general transcription system has a reason to know: your project code names, the internal tool everyone calls by an acronym, your colleagues' surnames, the author you keep citing, the framework you have opinions about. They are perhaps five percent of the words you speak and they appear in nearly every note.

Fixing these one at a time after every dictation is the thing that makes people quit. Smart Vocabulary is the fix: a personal dictionary with replacement rules, so a term you say constantly is spelled your way from then on. Add names as they come up rather than trying to seed the whole list in one sitting. Fifteen entries covers most of a working vault.

We wrote about how this compounds in custom vocabulary that learns your words. In an Obsidian context there is a bonus: consistent spelling of a proper noun is what makes links and unlinked mentions work at all. A name transcribed three different ways is three orphan concepts.

Editing what you dictated

Dictated text is a first draft with a specific error profile, not a mess. Expect the occasional homophone, an added or missing comma, and proper nouns landing wrong if you have not added them to your dictionary yet. What you do not usually get is the vague, cautious, half-committed prose that typing produces when you are editing while composing.

Draft in one pass without stopping, then repair in a second. Obsidian is a good place to do this because it is a fast editor with good keyboard navigation. Our guide to editing dictated text goes through the two-pass method in more detail, including using Voice Edit to speak a correction rather than hunting for the cursor position.

One habit worth adopting: read the note before you link it. A note that will be cited from ten other places should not contain a sentence you never actually looked at.

Other outliners work the same way

If you run more than one system, or you are still deciding, the approach transfers. Everything here applies with minor changes to dictating in Logseq and dictating in Roam Research, both of which are block-based rather than file-based but present the same basic surface: a text field, some syntax you should type, and a lot of prose you should not. We also have a combined walkthrough of voice dictation in Notion and Obsidian if you keep a foot in both.

A workflow that actually sticks

  1. Capture on the phone during the day. Ten-second dictations into the daily note. No structure, no links, no cleanup.
  2. Do the evening pass at the desk. Three or four spoken paragraphs about what happened and what you concluded.
  3. Once a week, open the stubs. Sort by created date, find the empty notes, and speak bodies into as many as you can in fifteen minutes.
  4. Link in a separate pass. Read the week's new prose, bracket the terms that deserve to be notes, and create the ones that do not exist yet.
  5. Add three vocabulary entries. Whatever got misspelled most this week.

The pattern underneath is that capture and structure are separate jobs. Voice is very good at the first and bad at the second, so stop making it do both at once.

Troubleshooting

Nothing appears when I speak. Focus is almost always the answer: the cursor is in the sidebar, the search field, or a modal. Click into the editor and check for a blinking cursor. If the hotkey itself is not firing, check Obsidian's Hotkeys settings for a conflict.

My note turned into chaos. Vim mode, normal mode. Undo, press i, try again.

Text goes into the command palette. The palette had focus when you started. Escape, click into the note.

Punctuation shows up where I did not ask for it. Trailing pauses at the end of a sentence are the usual cause. Finish the sentence and release cleanly rather than trailing off.

The same name is wrong every time. That is a vocabulary problem, not an accuracy problem, and adding one dictionary entry fixes it permanently.

A note on privacy

People choose Obsidian partly because the files are local and theirs. That instinct is worth respecting in the tools you attach to it. Voice Keyboard Pro's servers store only operational pings, no audio and no transcript content. Your notes stay in your vault, in plain text, on your disk.

The honest limitation on the other side: the transcription engine needs a connection, so dictation is not something you can rely on in a tunnel or on a plane without Wi-Fi. If offline capture matters to you, keep the built-in on-device dictation available as a fallback and use both.

Frequently asked questions

Do I need an Obsidian plugin to dictate?

No. A system-wide dictation tool types into the editor like a keyboard does. Nothing is installed in your vault and nothing is added to your synced settings folder.

Does this work in Obsidian Mobile?

Yes. On iPhone the microphone lives on the keyboard, so it works in Obsidian Mobile the same way it works in any other app.

Can I dictate wikilinks?

You can, but you should not. Speak the sentence and add the brackets afterwards, so the autocomplete can match against notes that actually exist.

Will dictated text break Markdown formatting?

Not if you keep syntax on the keyboard. Prose is prose in Live Preview and Source mode alike. The things that break are unbalanced brackets and stray backticks, which is exactly why you type those.

What about non-English notes?

Dictation supports multiple languages, and on iPhone two-way translation across 24 languages means you can speak in one language and have the note land in another. Machine translation is a solid first draft, not a final one, so have a person check anything that matters.

The bottom line

Obsidian rewards notes with actual content in them, and the reason most vaults do not have that is the four minutes of typing standing between a thought and a written paragraph. Speaking at 130 to 150 words a minute instead of typing at 40 does not make you a better thinker. It removes the tax that was quietly deciding which of your thoughts got written down.

Voice Keyboard Pro has a free tier with daily limits, and Pro is $4.99 a month or $34.99 a year. Try it on the stub notes first. Those are the ones where the difference is most obvious.