Faceless Desk
Production

How long should a faceless video script be?

A fixed speaking rate turns any runtime into an exact word count: 1,200 words is the mid-roll line, and a script that reads well often narrates badly.

Research-led · from vendor docs, not hands-on use 6 min read Updated 2026-09-04 Contains affiliate links

Disclosure: some links on this page earn us a commission if you subscribe, at no extra cost to you. It never changes what we recommend — where a free or cheaper tool wins, we say so.

A creator talking to camera can’t answer this question precisely. They ad-lib, pause, repeat themselves, and a 900-word outline becomes anything from five minutes to eleven.

Faceless video doesn’t work that way. The narration is synthesised at a rate you set, so the script length and the runtime are the same fact expressed in different units. That makes this one of the few production questions with an exact answer, and it’s worth using that precision rather than guessing and re-cutting.

The conversion

This site uses 150 words per minute, or about 900 characters per minute including spaces, as the working narration rate. It’s the same figure behind our narration cost breakdown, so the numbers on the two pages agree.

150 wpm is a deliberate choice, not a universal constant. Audiobooks sit around 150–160. Conversational speech runs faster. Most TTS engines default to roughly this pace and let you adjust it. If you speed your voice up, every number below shrinks proportionally — the arithmetic is easy to redo, which is why the rate is stated up front instead of buried.

The table

Target runtimeWordsCharactersGoogle WaveNetElevenLabselevenlabs.io · affiliate Creator
3 min4502,700$0.01$0.48
5 min7504,500$0.02$0.80
8 min1,2007,200$0.03$1.28
12 min1,80010,800$0.04$1.92
20 min3,00018,000$0.07$3.20

Narration costs come from the per-minute pricing breakdown. Note what they say: at the cheap end, script length is not a budget consideration at all. Doubling a video’s runtime costs three more cents. Your time writing it is the only real cost, which is worth remembering before optimising the wrong variable.

The 8-minute row is the one that matters

That row is bold for a reason. Videos of 8 minutes or longer are eligible for mid-roll ads; videos of 7:59 are not. It’s a hard threshold, and it’s the single biggest revenue difference any runtime decision makes — mid-rolls are widely reported to lift RPM substantially over pre-roll-only videos.

So there is a genuine cliff at 1,200 words. Below it you are leaving the format’s main monetisation lever unpulled.

The obvious move — pad everything to clear 8 minutes — is also the obvious trap. Padding costs retention, retention drives distribution, and a padded 8-minute video that loses half its viewers by minute three earns less than a tight 6-minute one that holds them. YouTube’s monetization requirements are watch-hour based, so filler works against you twice.

The honest version of the advice: pick topics with enough substance to reach 1,200 words without padding. That’s a decision made at the idea stage, not the writing stage. A thin topic stretched to 8 minutes is worse than the same topic told properly in 5, and much worse than a better topic chosen instead.

Why a good script can narrate badly

Word count is the easy half. The harder problem is that text written to be read and text written to be heard are different crafts, and most writing — human or machine — defaults to the first.

A reader controls the pace. They re-read the clause they lost, skim what they already know, and see the punctuation that shows how a sentence is built. A listener gets one pass at someone else’s tempo, with no ability to look back. Things that work fine on the page and fail in narration:

  • Long sentences with nested clauses. The eye tracks the structure; the ear loses the thread and can’t recover it.
  • Parenthetical asides. On the page they’re visibly set apart. Narrated, they’re indistinguishable from the main sentence.
  • Signposting that depends on layout — “as shown above”, “the second point” — when the listener never saw a first point written down.
  • Dense numbers. “$0.0036 per minute” is scannable in a table and near-meaningless spoken aloud. Say “about a third of a cent” and put the exact figure on screen.
  • Homophones and initialisms that a TTS engine will pronounce confidently and wrongly.

The reliable test costs nothing: generate the narration and listen to it before touching the visuals. Anywhere you have to concentrate to follow it is a rewrite, not a delivery problem. This is also the cheapest stage to iterate — at three cents a pass, re-rendering a script is free in practice.

Where AI writing tools actually fit

At the draft, not the final pass — and that’s a real limitation rather than a hedge.

General-purpose assistants and dedicated marketing writers alike are trained overwhelmingly on written prose, so their default output has exactly the properties listed above: balanced clauses, parentheticals, transitions that assume a reader. Asked for a video script, most will hand back an article with “[HOST]:” in front of it.

What they are genuinely good at is the part that isn’t voice: turning research into a structure, drafting a cold open, producing four variants of a hook to choose between, and getting a blank page to 1,200 words quickly. What they don’t reliably do is write for the ear. Prompting for it — “short sentences, one idea each, no parentheticals, written to be spoken aloud” — helps, and does not solve it.

The practical shape is AI for the structural draft, a human pass for the spoken rhythm, then the listen-back test. That also answers the tool question honestly: if you already pay for a general assistant like ChatGPT or Claude, you may not need a second subscription to a dedicated writer such as Writesonic for this specific job. The dedicated tools earn their place on volume, on team workflows and on templates for other formats — not on producing narration-ready prose out of the box.

The practical answer

  • 150 words per minute. Multiply your target runtime by 150 for the word count, by 900 for the character count.
  • Aim for 1,200+ words so the video clears 8 minutes and can carry mid-rolls — but only where the topic genuinely supports it.
  • Never pad to reach it. Choose a bigger topic instead; that is an idea-stage fix, not a writing-stage one.
  • Write for the ear: one idea per sentence, no nested clauses, no parentheticals, round the numbers aloud and show the precise ones on screen.
  • Listen to the render before editing visuals. It costs a few cents and catches everything the page hides.

Caveat: 150 wpm is a working assumption, not a rule — measure your own voice preset and scale every figure here proportionally. The narration costs are current vendor list prices and move without notice.

Sources