Short answer: Premiere Pro has no dictation button, but a system-wide Mac dictation tool types into any field you click. Use it for marker comments, clip logging notes, captions, and client emails. Set a hotkey that does not collide with Premiere's single-key shortcuts.
Open a project you finished three months ago and read your own markers. Most of them say something like "fix," "check audio," or "b-roll?" You knew exactly what you meant when you typed it. You do not know now.
That is not carelessness. It is arithmetic. You are in the middle of a cut, the thought is complete in your head, and typing out "the room tone changes at the cut because the second interview setup moved the mic, so this needs a crossfade rather than a hard cut" costs about twenty seconds and pulls both hands off the keyboard shortcuts you were using. Typing "fix audio" costs two seconds. So you type "fix audio" and lose the thought.
Speaking that same sentence takes about six seconds. The economics flip, and the note that survives to your next session is the one that actually says something. This guide covers how to dictate in Premiere Pro on a Mac: what works, what to type instead of speak, and the one setup mistake that will make your timeline jump around the first time you try it.
What Premiere's built-in transcription does, and why it is a different job
Premiere Pro has speech-to-text features. It can transcribe the dialogue in your sequence, generate captions from that transcript, and let you search spoken words in the Text panel. These are genuinely useful, and if you cut interview or documentary work you probably already use them.
They solve a different problem. Premiere's transcription takes audio that already exists in your timeline and turns it into a searchable record of what other people said. It does not put your words into a text field. When your cursor is sitting in a marker comment box, or a clip's Description column, or the subject line of an email to a client, Premiere's transcription is not involved at all. That is the writing that eats your day, and it is all still keyboard work.
So there are two separate needs here, and conflating them is why people search for this and come up empty:
- Transcribing footage — the words your subjects said, timed to the timeline. That is the Text panel's job.
- Writing your own text — notes, logs, titles, feedback, emails. That needs something that types at the cursor.
The rest of this guide is about the second one.
Nothing gets installed in Premiere
This matters more to editors than to most people. Post-production Macs are fragile in a specific way: a plugin that misbehaves can take down a session in the middle of a deadline, and a studio that manages your machine may not let you add anything to the Adobe application folder at all.
Voice Keyboard Pro is a menu bar app. It does not install a Premiere extension, it does not touch your Creative Cloud account, and it does not add anything to your project or your media. It types text into whatever field on your Mac currently has the cursor, exactly as if the keyboard produced it. Premiere has no idea anything unusual happened. That also means there is nothing to reinstall after a Premiere update, and nothing for a facility IT team to approve inside Adobe.
The workflow is: hold a hotkey, speak, release, text appears where you were typing.
Setup: the hotkey collision problem
Premiere Pro has one of the densest single-key shortcut maps of any Mac application. J, K, and L are shuttle. I and O are in and out. M sets a marker. Q and W are ripple trim. Almost every letter does something, and many of them do it the instant you press the key.
This is the mistake that ruins the first attempt. Someone picks a comfortable dictation hotkey, presses it while the timeline has focus, and watches Premiere ripple-delete a section or jump the playhead somewhere unexpected. Two rules prevent it.
1. Click into the text field first
Premiere routes keystrokes by panel focus. If the timeline is focused, keys are commands. If a marker comment box or a metadata cell is focused, keys are text. Before you speak, click directly into the field you want the text to land in and confirm you can see a blinking cursor. This is the single most common cause of "I dictated and nothing appeared" in any editing application.
2. Pick a hotkey Premiere does not want
Use a key that is not part of Premiere's editing vocabulary. Good options:
- A function key (F13 through F19 if your keyboard has them, or a remapped F-key you do not use for anything else)
- A lone right-side modifier held on its own
- A key on an external control surface or Stream Deck mapped to your dictation hotkey
Then verify it. Open Premiere's Keyboard Shortcuts editor (Premiere Pro → Keyboard Shortcuts), type your chosen key into the search field, and confirm nothing is bound to it. Do this once, at the start, and you never think about it again. If you also use Final Cut Pro or DaVinci Resolve, pick one hotkey that is clear in all three so your hands learn a single habit.
The five places dictation actually pays off
1. Marker comments
This is the highest-leverage surface in the whole application, and it is the one that gets the worst treatment.
Markers are where an edit talks to its future self. They are also where the cost of typing is highest, because you set them mid-flow with the playhead parked and your hands on shuttle keys. The result is the two-word marker that means nothing later.
A dictated marker sounds like a sentence because it is one. Try a consistent three-beat shape: what is wrong, why it is wrong, what to do about it. "The cutaway at this point lands too early, the line before it has not finished its thought, so push it four frames later and see if the reaction still reads." That is one breath. It is also a note that a second editor, or an assistant, or you in October, can act on without watching the surrounding two minutes to reconstruct your reasoning.
The same applies to markers you are leaving for someone else. Notes to a colorist, a sound mixer, or a graphics artist are handoff documents. Handoff documents written in fragments generate a round of clarifying messages. Handoff documents written in sentences do not.
2. Clip notes and the logging pass
Logging is the part of the job everybody agrees matters and almost nobody does properly, because it is pure typing with no creative payoff. You have a bin of ninety clips, and the Description, Comment, and Log Note columns are all empty.
Dictation changes the shape of this task. Scrub a clip, click into its Description cell, speak one sentence about what is in it and whether it is usable, tab to the next. You are describing what you are already looking at, which is the easiest kind of speech there is. A logging pass that would take a long afternoon of typing becomes something you can get through while the footage plays.
The payoff arrives later, when search actually returns something. Bins full of one-word logs are not searchable in any meaningful sense. Bins full of sentences are.
3. Captions and subtitle text
Here the division of labor is specific. Premiere's own transcription should generate your first pass of dialogue captions. That is what it is for, and speaking your subjects' words back at your Mac would be slower and less accurate than letting the transcript do it.
Where dictation helps is the caption text that is not dialogue, and the corrections. Forced narrative titles, on-screen text you are transcribing for accessibility, sound-effect captions, and speaker identifications are all things you write rather than things anyone said. So are the rewrites when a caption needs to be shortened to fit a reading speed. Click into the caption text field, speak the corrected line, move on.
The timing stays keyboard and mouse work. Dictation writes words. It does not drag caption boundaries.
4. Titles, lower thirds, and graphics text
Essential Graphics text layers accept dictated text the same as any other field. This is genuinely useful for longer on-screen copy: pull quotes, disclaimers, credits blocks, explainer captions, the paragraph of context that goes under a chart.
One warning that is worth more than the time it saves. Do not dictate a person's name into a lower third and move on. Names are exactly where transcription is most likely to guess, and a misspelled name in a lower third is the error that reaches the client, then the audience, then your inbox. Speak the sentence, then look at the name with your eyes before you approve the graphic. If you cut for the same client repeatedly, there is a better fix, covered below.
5. The writing that happens around the edit
Count honestly and this is probably the biggest category. The email explaining what changed in version four. The note in a review platform answering a comment. The delivery summary. The YouTube description and the chapter list. The scope conversation about whether the extra deliverable is in the estimate.
None of this is in Premiere, and all of it is the reason your edit day runs past the edit. Because dictation types at the cursor system-wide, the same hotkey works in your email client, your browser, your notes app, and your invoicing tool. This is also where the speed difference is most obvious: an average adult types around 40 words per minute and speaks at 130 to 150, and a three-paragraph client email is exactly the length where that gap turns into real minutes.
Speak the sentences, type the values
This is the rule that separates editors who stick with dictation from editors who abandon it in week one. Spoken numbers and identifiers fail quietly. A wrong word in a sentence looks wrong. A wrong digit in a timecode looks completely normal, and gets acted on.
Type these, always:
- Timecode. Speaking "01:14:22:07" invites a transposition you will not see, and the note will send someone to the wrong frame.
- Frame rates, resolutions, bit depths, and codec names. These are strings, not prose.
- Effect parameters and values. Scale, opacity, keyframe numbers, LUT names.
- File names, sequence names, and version numbers. "v04_final_rev2" is not a sentence and should never be spoken.
- Client-supplied job or PO numbers. Same reason.
Speak everything that is prose: the description, the reason, the instruction, the context, the paragraph explaining what you tried. That is where all the volume is anyway. The numbers are a handful of characters; the explanation around them is a hundred words.
A wrong word in a sentence is visible. A wrong digit in a timecode is invisible until someone conforms to it.
Names are the accuracy problem, and there is a fix
Every project has a vocabulary that no general transcription has ever encountered: the client's company, the product line, the talent's surname, the internal name of the campaign, the abbreviation your team uses for a deliverable. These words are maybe five percent of what you say and appear in nearly every note you write, so a miss on them is a correction on almost every line.
Smart Vocabulary is the answer to this. It is a personal dictionary with replacement rules: you add the spellings that matter to you, and they come out right without you fixing them by hand each time. Add the client, the product names, the talent, and the two or three abbreviations your team uses, and the correction rate on your notes drops sharply. Set it up once at the start of a project, add to it when something new appears, and you stop noticing the problem. Our guide to building a custom vocabulary that learns your words covers the approach in more detail.
Client review calls
Review calls are a special case because you are not writing during them, you are listening and nodding, and afterward you have to reconstruct twelve pieces of feedback from memory.
Meeting Mode on the Mac app handles the capture side. It detects speakers and produces AI notes, so you come out of the call with a record of what was asked for rather than a memory of it. Calendar meeting detection means it can be ready when the call starts instead of after you remember to open something.
Two boundaries, both real. Those notes are raw material, not a revision list: read them, decide what is actually a change request, and rewrite them into the numbered list you will actually work from. And tell the people on the call that you are taking notes with an assist. Recording and note-taking norms differ by client, jurisdiction, and contract, and the person on the other end deciding they were captured without knowing is a much bigger problem than the twenty minutes you saved. If your client work is under an NDA that governs how project material is handled, check it before you route anything about the footage through any external service. Our Mac meeting transcription guide goes deeper on how this fits a working day.
Fixing what comes out
Dictated text is a first draft, and treating it as finished is how people conclude dictation "does not work." Two tools shorten the cleanup.
Voice Edit lets you speak a change rather than hunting for the cursor position. If a marker note came out with the wrong emphasis or a run-on clause, saying what should change is faster than clicking into the middle of a sentence in a small comment box. On the iPhone keyboard this matters even more, because precise cursor placement on a phone is genuinely unpleasant.
The other habit is a two-pass rhythm: speak the whole note without stopping, then read it once and fix. Stopping mid-sentence to correct a word is what makes dictation feel slow, because you lose the thought you were holding. Our post on editing dictated text lays out the workflow.
Notes from the phone
Ideas about an edit rarely arrive at the workstation. They arrive on the walk, in the car before you go in, or at 11pm when you realize why the second act is not landing.
The iPhone keyboard has a mic button built in, so it dictates into whatever app you use for notes, into a message to your producer, or into an email to yourself. Speaking a paragraph into your phone takes about as long as it takes to think it. The note is in your inbox when you sit down, and it says the whole thought rather than three words that were supposed to remind you of it.
Working with international clients
If you deliver to clients who work in another language, two-way translation while dictating covers 24 languages, so you can speak in English and produce a note or an email in the client's language.
Use it the way you would use any machine translation: as a strong first draft. It is very good for a status update or a note about a cut. Have a human check anything contractual, anything about payment, and anything that would be expensive to get subtly wrong.
Honest limits
Some things dictation does not do, and pretending otherwise wastes your time:
- It does not edit. Trimming, timing, color, and audio work are hands-on. Voice writes text; it does not cut a sequence.
- It needs a connection. Transcription runs on our cloud infrastructure. On a plane or in a machine room with no network, keep macOS's built-in dictation as a fallback for short notes.
- A loud room costs accuracy. Edit suites are usually quiet, which helps, but monitor audio playing through speakers while you speak will degrade results. Pause playback, or use headphones.
- It does not make you a better editor. It removes the friction between having a thought and recording it. The thought still has to be good.
- Dense timeline sessions stay keyboard work. When you are trimming for an hour with no writing in between, dictation is simply not involved.
A workflow that fits an edit day
- Before the project starts: add the client, product, and talent names to Smart Vocabulary. Two minutes.
- During the logging pass: click into Description, speak one sentence per clip, tab down.
- During the cut: set the marker with your normal key, click into the comment field, speak the full note, keep going.
- After a review call: turn the captured notes into a numbered change list in your own words.
- End of session: dictate the client update while the session is still in your head, not the next morning when it is not.
Frequently asked questions
Does this work with Premiere Pro on Windows?
The Mac app is macOS only. Premiere runs on both platforms, but the menu bar app described here does not have a Windows version.
Will it interfere with Premiere's own speech-to-text?
No. They operate on different things. Premiere's transcription reads audio in your sequence. Dictation types characters into a focused text field. They can be used in the same session without either one knowing about the other.
Can I dictate into the Text panel to correct a transcript?
Yes, if you can click into the editable text and see a cursor. Transcript corrections are one of the better uses of it, since you are usually fixing a proper noun or a misheard phrase and typing it out is fiddly.
Should I dictate expressions or scripting?
No. Anything with brackets, exact syntax, or case sensitivity should be typed. The rule is the same as with timecode: if it is a string rather than a sentence, use the keyboard.
What happens to what I say?
Audio is transcribed and the text is returned to your cursor. Our servers store only operational pings, not audio and not the content of what you dictated. If your client work is covered by an NDA, review it yourself and make your own call about what is appropriate to route through any external service.
Do I need to pay for it?
There is a free tier with daily limits, which is enough to find out whether marker notes and logging feel different. Pro is $4.99 a month or $34.99 a year and removes the limits.
The test worth running
Take one project. For a week, dictate every marker comment instead of typing it, and log one bin by voice. Then open the project cold the following week and read your own notes.
The difference will not be in how fast you cut. It will be in whether your markers still make sense, and in how much of the next session you spend reconstructing what past-you meant. Editors who work on long projects, or who hand off to anyone, tend to notice that first.
Voice Keyboard Pro has a free tier if you want to try a logging pass before deciding.