A desk with a microphone arm, a laptop and a ring light switched off

Use cases

Creators: you write more than you film

The camera work is the visible part. The unglamorous volume is scripts, descriptions, captions, newsletters, sponsor emails and replies — and that is where the week actually goes.

Last updated

Scripts are the obvious one and people still type them

A script exists to be spoken. Writing it by typing means composing in one register and delivering in another, and the join shows — that slightly stiff quality in the first thirty seconds of a lot of videos is a typed script being read.

Dictating it removes the translation. You say the thing, roughly, the way you would say it on camera. The transcript is the script, already in your voice, already at the right rhythm. Then you tighten it.

  1. Talk through the video once, loosely Not performing. Explaining it to one person. Three to five minutes.
  2. Read the transcript The good lines are already in there, usually in the second half where you stopped thinking about it.
  3. Cut to the structure Most of it goes. What is left is in your voice and needed no rewriting to get there.
  4. Dictate the hook separately Say ten versions of the first fifteen seconds, back to back. Choose one. This is where the video is won or lost.

The volume nobody counts

TaskPer video or per weekDictate?
ScriptPer videoYes — the biggest single win
DescriptionPer videoYes
Chapter summariesPer videoYes
TitlesPer videoNo. Compression, not writing
Thumbnail textPer videoNo
Pinned commentPer videoYes
Comment repliesDozens a weekYes — the compounding one
NewsletterWeeklyYes
Sponsor emails and briefsWeeklyYes
Community postsWeeklyYes

The pattern: everything that is prose is worth speaking; everything with a character limit is worth typing. Titles and thumbnail text are compression tasks and speaking makes them longer, which is the wrong direction.

Replies are the compounding activity

Most creators know that replying properly to comments builds an audience faster than anything else they do, and most reply with three words because typing forty replies is an hour.

Fifteen seconds of speaking per reply changes the arithmetic. Forty specific replies in ten minutes rather than an hour is the difference between doing it and intending to.

Why on-device matters for creators specifically

Two reasons that are not the usual privacy argument.

  • Unreleased ideas. Your script is your next video. Dictating it into a cloud service means your unpublished content sits on someone else's infrastructure before it exists. Most creators would not email a draft to a stranger and this is closer to that than it looks.
  • Sponsor and brand material under NDA. Rates, deal terms, unannounced campaigns and embargoed products come up constantly, and brand agreements are usually written broadly enough that a third-party processor is a problem.

There is also a practical one: creators travel, and on-device dictation works on the plane where your script was going to get written anyway.

Setting it up

  1. Teach it your own vocabulary Your channel name, your recurring segments, the products you cover, the people you collaborate with. A creator's writing is unusually dense with proper nouns.
  2. Make snippets for the fixed parts Your standard description block, your affiliate disclosure, your sponsor enquiry reply, your rates paragraph.
  3. Dictate the loose pass, then cut Not the finished script in one go. Surplus then subtraction.
  4. Keep titles on the keyboard Every time.
  5. Use a close microphone You probably already own a good one. Use it for dictation too — accuracy tracks the signal.
Snippets in VV Your description block, your disclosure and your rates paragraph, from three spoken words.

Questions

Will a dictated script sound more natural?

Usually, yes — it was spoken rather than composed, so it already has your rhythm. The trade is that it will be longer and looser than a typed script, so budget an editing pass for cutting rather than rewriting.

Does it work in YouTube Studio and Substack?

Yes, in any browser, with no extension. A system-wide dictation app types where your cursor is.

Should I dictate titles?

No. Titles are compression and speaking produces expansion. Dictate fifteen rough directions if it helps you think, then write the title at the keyboard.

Is my unreleased script safe?

With on-device recognition the audio is not transmitted anywhere, so your unpublished content stays on your Mac. With a cloud dictation service it is sent to that vendor as you speak it.