How to Set Video Length in Synthesia for Marketing Video

Why Runtime Is a Strategy Decision, Not a Formatting Detail

Most teams treat video length as something that happens to them. They write a script, paste it into an AI avatar tool, hit generate, and discover the result is 2 minutes 47 seconds when the campaign needed 45 seconds. Then they start cutting words in a panic, the pacing collapses, and the CTA lands on top of a sentence that no longer makes sense.

The teams that ship reliably do the opposite. They decide the runtime first, then write into that box. Runtime is not a cosmetic detail — it directly controls completion rate, cost per view, ad delivery, and whether your call to action ever gets seen.

Three things make length strategically important:

Retention is non-linear. Most viewers leave in the first few seconds and again at natural scene boundaries. A video that is 20 seconds too long does not just lose 20 seconds of attention; it loses the percentage of the audience who drift before the close.
Platforms reward completion. Feeds that rank on watch-through will push a tight 40-second clip harder than a padded 90-second one, even when the topic is identical.
Production cost scales with length. Every additional 30 seconds means more script passes, more avatar renders, more audio checks, and more review cycles from legal and brand.

The practical consequence: your script and your runtime are the same decision, expressed in two different units. Learn to convert between them and the rest of the workflow becomes predictable.

The Script-to-Duration Formula You Should Memorize

AI avatar video is unusually forgiving about predicting length, because the voice is synthetic and consistent. A human presenter speeds up when nervous and slows down when reading numbers. A synthetic voice does exactly what its rate setting says.

Words per minute is your only real variable

The core formula:

seconds = (word_count / words_per_minute) * 60

Rearrange it when you have a hard limit:

target_words = (seconds / 60) * words_per_minute

Most AI presenter voices land between 140 and 170 words per minute at default settings. Conversational marketing copy sits comfortably at 150. A calm corporate narration sits at 135–145. An energetic social voice sits at 165–180.

A quick reference table

Assuming 150 words per minute, which is a good planning default:

Target length	Script budget
15 seconds	~38 words
30 seconds	~75 words
45 seconds	~113 words
60 seconds	~150 words
90 seconds	~225 words
2 minutes	~300 words
3 minutes	~450 words
5 minutes	~750 words

Write these numbers on a sticky note. When a stakeholder asks for a 30-second video and hands you a 300-word script, you can show the arithmetic instead of arguing about taste.

Subtract a buffer for pauses, numbers, and branding

Raw word count underestimates runtime, because speech is not a continuous stream of syllables:

Sentence boundaries add roughly 0.3–0.6 seconds each in a natural delivery.
Read-out numbers are longer than they look. $1,499 is four spoken words. Q3 2026 is five syllables.
Acronyms and product names take longer than their character count suggests, especially unusual brand names.
Intros, logos, and end cards are usually not script words at all but still consume seconds.
Breath and emphasis pauses are inserted automatically by most tools and are not visible in your text editor.

A reliable habit: after estimating, add 8–12% for a 60-second video and 12–18% for anything under 30 seconds, where a single extra sentence is proportionally huge.

Work backward from the hard limit, not forward from the draft

If you have a 45-second slot and a 150 wpm voice, the math gives you about 113 words. Because of pauses and the end card, budget 95–100 words of actual narration. That is a short paragraph and a half. Knowing that before you write saves an entire draft cycle.

Matching Runtime to Marketing Objective

Length should follow from what the viewer is supposed to do next. A common mistake is choosing a duration because it "feels" right, then trying to cram a mismatched objective into it. Start with the job.

Use this for a single message delivered to a cold audience. One problem, one promise, one action. You have room for a hook, two supporting lines, and a CTA — nothing else. If your script includes the phrase "and also," cut what follows.

Good objectives: product awareness, event reminder, feature spotlight, retargeting teaser.

30–60 seconds: explainers, retargeting, and demo teasers

This is the workhorse range. You can introduce a problem, show one product mechanism, and close. It is long enough to feel credible and short enough that completion rate stays healthy on most feeds.

Good objectives: cold-traffic explainer, feature launch, comparison against an alternative, lead magnet promotion.

60–120 seconds: product tours, onboarding, and objection handling

Once you pass a minute, viewers have self-selected as interested. Now you can structure: context, three beats, proof, CTA. Keep each beat under 25 seconds so the visual rhythm changes often enough to hold attention.

Good objectives: onboarding sequence, sales follow-up asset, mid-funnel objection handling, support deflection.

2–5 minutes: training, deep dives, and enablement

At this length you are producing a reference asset, not a feed asset. Structure matters more than polish. Add chapter markers, on-screen section titles, and a summary in the final 20 seconds because many viewers scrub rather than watch linearly.

Good objectives: internal training, detailed product walkthrough, compliance or policy briefing, recorded webinar follow-up.

A Repeatable Workflow for Locking Duration in Synthesia

The following sequence works in any AI avatar video tool, including Synthesia, and keeps runtime predictable from the first draft.

Step 1: Fix the target runtime before opening the editor

Write the number down: 30 seconds, 75 seconds, 3 minutes. Everything downstream is measured against it. If different stakeholders want different lengths, produce two scripts rather than one compromised middle length.

Step 2: Convert the runtime into a word budget

Use the formula. Then apply the buffer. This gives you a maximum word count you should not exceed without a conscious decision.

Step 3: Choose the avatar and voice before finalizing the script

Voice selection changes the mathematics. Switching from a calm narrator at 140 wpm to a bright presenter at 175 wpm can cut 20 seconds from a two-minute script. Pick the voice early, note its approximate rate, and re-check your word budget against it.

Step 4: Write in timed blocks, not paragraphs

Divide the script into scenes, each with its own runtime target. A 60-second video might break down as:

Hook: 5 seconds (~12 words)
Problem: 12 seconds (~30 words)
Solution: 20 seconds (~50 words)
Proof: 15 seconds (~38 words)
CTA: 8 seconds (~20 words)

Timed blocks make trimming surgical. When the render comes back 6 seconds long, you know which block to attack and by how much.

Step 5: Use the timeline and scene structure to control pacing

In Synthesia, each scene is a block of script plus a visual. Splitting a long scene into two shorter ones lets you change background, camera framing, or on-screen text, which reads as a pacing change even if the voice rate is identical. Pacing is what viewers perceive as energy, and scene boundaries are the cheapest lever you have.

Step 6: Reconcile audio and lip-sync across the whole runtime

Synchronization problems usually appear at the extremes: the very first second, when audio can clip before the avatar settles into the animation, and the very end, when the last word can get cut. Generate the full video and listen with headphones at the start, at each scene transition, and at the close. If a word is clipped, add a fraction of a second of padding rather than re-recording the whole scene.

Step 7: Export, measure, and iterate on the real numbers

The editor estimate is a planning tool. The exported file is the truth. Compare actual runtime against your prediction and record the delta. After three or four projects you will have a personal wpm figure for your preferred voice, and your estimates will land within a second or two.

Techniques for Enforcing a Hard Time Limit

When you are 15% over and the deadline is real, these are the levers in order of least damage.

Trim the script, not the pacing

Cuts should remove whole ideas, not individual words. Deleting adjectives produces a breathless sentence that is barely shorter. Removing an entire supporting sentence usually saves 4–6 seconds and often makes the argument sharper.

Cut the second example

Marketing scripts love parallel examples. One is usually enough. The second example is the single most common source of 8–10 seconds of fat.

Move the CTA earlier

If you are consistently over budget, the CTA is landing too late. Put the action at the 70% mark and let the final seconds reinforce it rather than introduce it.

Adjust speaking rate — carefully

Most tools let you speed up or slow down delivery. Small increases (up to about 10%) are usually undetectable. Beyond that, synthetic speech starts to sound rushed and unpersuasive, and the lip-sync has to work harder. Use rate as a fine adjustment, never as a substitute for editing.

Let visuals carry information

A list of four benefits takes 12 seconds to say and 3 seconds to read. Move enumerations to on-screen text or a simple animation, and speak only the one benefit that matters most.

Split rather than compress

When the content genuinely needs 3 minutes but the slot is 60 seconds, produce a series instead of a compressed single. Three focused clips almost always outperform one overloaded one, and you get three assets for testing.

Platform-by-Platform Duration Rules of Thumb

Vertical short-form feeds

Aim for 21–34 seconds for cold audiences. This range is long enough to deliver a complete thought and short enough to loop, and looping multiplies your effective watch time. If your content needs more room, structure the first 3 seconds so they work even without sound.

Long-form video platforms

Mid-length content (4–8 minutes) performs well when the topic is searchable and the viewer arrived with intent. For AI avatar content, the risk is visual monotony rather than length, so change scene every 15–25 seconds and add screen recordings or slides where the narration would otherwise carry everything.

Website embeds and sales pages

Keep autoplay-muted embeds under 90 seconds. Sales pages where the viewer clicked play can support 2–3 minutes because intent is already established. Place the primary CTA button below the player and repeat the same action in the video's final 5 seconds.

Paid ads

Assume a meaningful share of viewers will skip at the earliest opportunity. Front-load the value proposition into the first 3 seconds and place the brand mention inside the first 5. For skippable formats, a 30–45 second cut usually gives the best balance between message completeness and cost efficiency.

Common Mistakes That Blow Up Runtime

Writing the script before choosing the voice. The voice sets the speed. Choosing it last means rewriting for time.
Ignoring pauses in the word count. A script with 40 short sentences runs far longer than 150 words per minute suggests.
Budgets with no end card allowance. Logos, legal lines, and disclaimers can eat 4–8 seconds you never planned for.
Scene bloat. A scene that changes nothing visually but runs 40 seconds feels twice as long as it is.
Adding a second CTA. Two asks split attention and generally add 5 seconds each.
Localizing without re-timing. The same script is rarely the same length in another language.
Trusting the editor preview over the export. Small discrepancies compound across scenes.
Never measuring. If you do not record actual runtime against predicted runtime, you will repeat the same estimate error indefinitely.

Localization: Duration Shifts When You Change Language

If your video will be translated or dubbed, length control becomes a localization problem. Text expansion is real and substantial:

Romance languages such as Spanish, French, Italian, and Portuguese typically expand 15–25% over English.
German frequently expands 20–30%, especially with compound nouns.
Polish tends toward longer phrasing and more syllables per word than English.
Japanese and Simplified Chinese often use fewer characters but similar or slightly longer spoken durations, because each character carries multiple syllables. Do not assume a character-count estimate is accurate for either.

Practical implications:

Build your English script to about 85% of the target runtime if you know localization is coming.
Keep sentences short. Long subordinate clauses expand unpredictably.
Avoid idioms and wordplay, which either expand or lose meaning.
Re-check lip-sync after dubbing; the avatar's mouth movements are generated per language and timing can shift by a few frames.
Test one scene in every target language before committing to the full production.

QA Checklist Before You Publish

Run this list every time. It takes four minutes and prevents most reshoots.

Runtime matches the target within 1 second.
First 3 seconds contain the hook, even muted.
Every scene has a visual change or on-screen text.
No sentence is clipped at a scene boundary.
Audio levels are consistent from the first sentence to the last.
Lip-sync checked at the opening and closing seconds.
CTA appears in speech and on screen.
Captions are synced and do not overflow the safe area.
Brand colors, logo placement, and legal text are correct.
Exported file plays correctly on a phone at 1x speed.
File naming follows your asset convention so it is findable later.

FAQ

How do I set an exact video length in Synthesia?

Synthesia builds runtime from your script, so there is no single "set length to 45 seconds" box. Control it in two places: the script itself, using a word budget derived from your voice's words-per-minute rate, and the scene structure, which determines pauses and transitions. Then verify with the timeline estimate and confirm with the exported file.

How many words is a 60-second AI avatar video?

At 150 words per minute, roughly 150 words. Because of pauses, numbers read aloud, and an end card, budget 130–140 words of narration if you need to land close to exactly 60 seconds.

Can I speed up the voice to fit more script?

Yes, but keep increases under about 10%. Beyond that, delivery sounds rushed and less credible, and lip-sync accuracy degrades. Speeding up is a fine adjustment, not a way to fit 90 seconds of content into 60.

Why does my video come out longer than the editor estimated?

Usually punctuation. Many short sentences each add a fraction of a second of pause. Numbers, acronyms, and unusual product names also take longer to speak than their character count implies. Add a 10% buffer to your estimate and the gap closes.

Should I make one long video or several short ones?

Match the format to the objective. Cold audiences need short clips because attention is unproven. Post-purchase audiences and internal viewers tolerate longer content because intent is already established. When in doubt, produce the short version first and add a longer cut only if data supports it.

How often should the visual change to keep pacing strong?

Roughly every 15–25 seconds for long-form content and every 5–10 seconds for short-form. The change can be modest — a new background, a camera angle shift, an on-screen bullet — as long as something moves.

Does video length affect cost?

Indirectly, yes. Longer videos take more script revisions, more render time, more review cycles, and more localization work. The biggest hidden cost is iteration: estimating runtime accurately the first time removes entire draft cycles from your production schedule.

Putting It Together

Runtime control is a habit more than a feature. Decide the length, convert it to a word budget, pick the voice before you write, structure the script in timed blocks, and verify against the exported file. Do that consistently and the question stops being "why is this video 20 seconds too long" and becomes "which of these two 40-second cuts should we run."

Start with one small change on your next project: write the target runtime at the top of the document and the word budget next to it. Everything after that gets easier.