How to Set Video Length in Synthesia for Marketing Spots

Most people assume the hard part of an AI avatar video is the script. In practice, the hard part is time. A script that reads beautifully on a page can land at 71 seconds when the ad slot needs 60, and the difference between those two numbers is often the difference between a video that gets watched to the end and one that gets scrolled past in the first three seconds.

Synthesia gives you a surprising amount of control over runtime, but almost none of it is labelled "set video length." Duration is an emergent property of several decisions: how many words are in the script, how fast the avatar speaks, how long the pauses between scenes run, how many scenes you stitch together, and what resolution and frame rate you export at. Getting predictable length means understanding how those levers interact, then building a workflow that treats runtime as a spec rather than an afterthought.

This guide walks through that workflow from start to finish: how to plan a target duration, how to edit a script to hit it, how to pace scene by scene, how to tune voice and pause settings, and how to fix the most common length problems when a render comes back wrong.

How Synthesia Determines Duration by Default

When you paste a script into a Synthesia scene and generate, the platform estimates duration from the character count of the text and the speaking rate of the selected voice. That estimate is usually close, but it is not a lock. A few things move it.

Script length is the invisible clock

Roughly, an English-speaking avatar voice reads at about 130 to 160 words per minute. A casual, warm voice sits at the low end; a brisk corporate narrator sits at the high end. That means a 180-word script lands somewhere between 68 and 83 seconds depending on voice choice alone. If you write to a word count without choosing a voice first, you are aiming at a moving target.

Punctuation and line breaks count

Commas create micro-pauses. Periods create longer ones. A paragraph break can add a beat of silence. This is why two scripts with identical word counts can differ by five to eight seconds. Short, punchy sentences — the kind that work well in ads — actually add time because each full stop becomes a small pause.

Scene count multiplies the drift

Synthesia videos are usually built from multiple scenes. Each scene boundary introduces a transition and a small gap. Ten scenes might add three to six seconds of non-speaking time across the video. If you are targeting exactly 30 seconds, that overhead is more than 15 percent of your budget.

Decide Your Target Length Before You Write a Word

The single biggest cause of painful re-editing is writing first and measuring later. Flip the order.

Start from the placement, not the idea

Ask where the video will actually run before the script exists. A TikTok hook, a LinkedIn feed video, a homepage hero, a YouTube pre-roll, and an event loop all have different natural lengths. Pick one primary placement and treat every other placement as a derived cut.

A simple planning grid:

• 6 to 10 seconds: single-idea bumper, one avatar line, no scene changes.
• 15 seconds: one problem, one product moment, one call to action.
• 30 seconds: problem, solution, proof, CTA — the workhorse ad length.
• 60 seconds: adds a second proof point or a short demo.
• 90 seconds to 3 minutes: explainer or onboarding, best for owned channels.

Convert the target into a word budget

Once you know the target and the voice, calculate a budget. At 145 words per minute, that is about 2.4 words per second.

• 15 seconds → roughly 36 words of speech, minus overhead → aim for 30 to 34 words.
• 30 seconds → roughly 72 words of speech → aim for 62 to 68 words.
• 60 seconds → roughly 144 words → aim for 130 to 138 words.

Write the budget at the top of your script document and check the count as you draft. It is far easier to cut a sentence while writing than to rebuild a finished render.

Editing a Script to Hit a Runtime

When a draft runs long, most people delete whole sentences. That is often the wrong cut, because it removes meaning instead of removing friction.

Cut with these five passes, in order

1. Delete throat-clearing. "In today's fast-moving world, businesses need..." becomes "Businesses need..."
2. Replace clauses with verbs. "What we do is help teams to reduce..." becomes "We cut..."
3. Kill duplicated ideas. Ads frequently state the same benefit twice with different wording. Keep the sharper version.
4. Compress numbers and lists. Three examples become the strongest one.
5. Tighten the CTA. "If you'd like to learn more about how we can help, click the link below" becomes "Start free — link below."

Add time deliberately when you are short

Being under target is usually easier to fix than being over, but do not pad with adjectives. Add one of these instead, in order of usefulness:

• A second proof point (a number, a customer outcome, a comparison).
• A short objection-handling line ("No setup, no migration.").
• A held final shot with the logo and CTA on screen while the voice finishes.

A held end card is the cheapest way to add three to five seconds without writing anything new.

Scene-Level Pacing: The Real Control Surface

Synthesia's scene structure is where pacing actually lives. A video is not one continuous take; it is a sequence of cards, each with its own script, avatar, background, and layout. That structure is a gift for pacing if you use it as a storyboard rather than a text container.

Assign a time budget per scene

Before generating anything, write the intended seconds next to each scene in your script document. Then verify against the real render. A 30-second ad might look like this:

Scene	Purpose	Target
1	Hook question or stat	4s
2	Problem in one line	6s
3	Product moment with visual	9s
4	Proof or result	7s
5	CTA with end card	4s

That is 30 seconds including transitions. If scene 3 renders at 12 seconds, you have found the problem early rather than at the end.

Change scenes to change rhythm

Pacing is not just total duration; it is the rate at which visual information changes. A 60-second video with one static avatar shot and one background feels longer than a 60-second video with five distinct visual beats. When a video feels sluggish at the right length, the fix is usually more scene changes, not fewer words.

Conversely, if a video feels frantic, consolidate scenes and let one visual breathe for eight to ten seconds. Long holds read as confidence in a premium ad; rapid cuts read as urgency in a performance ad. Choose based on the message.

Match scene length to sentence length

A useful micro-rule: keep each scene's script to one idea and one to three sentences. If a scene needs four sentences, it probably should be two scenes. This keeps your ability to reorder, swap, or localize without breaking flow.

Transitions, Pauses, and Silence as Pacing Tools

Silence is the most underused lever in avatar video. Two levers control it directly.

Voice speed

Most avatar voices expose a speaking-rate adjustment. Slowing a voice by 10 percent can add six to eight seconds to a one-minute script while also improving comprehension for complex or technical content. Speeding up by 10 percent can pull a 66-second cut under the 60-second mark without deleting a single word — though it starts to sound rushed above roughly 115 percent.

Use speed adjustment as a finishing tool, not a planning tool. Write to the natural pace, then nudge.

Pauses and transition types

Transitions between scenes behave differently on the timeline. A hard cut feels instant; a fade or slide adds perceived time. If you are within two seconds of target, switching three transitions from fade to cut can be the entire fix. Conversely, if a video feels choppy, longer fades create smoother pacing without changing a word.

You can also build deliberate pauses by ending a scene with a short line and letting the final frame hold. A one-second hold before the CTA is one of the highest-impact edits in short-form advertising — it gives the viewer a beat to register the offer.

Platform Targets and the Lengths That Work

Runtime only matters relative to the platform's behavior. The same script should be exported as several lengths.

Short-form vertical (TikTok, Reels, Shorts)

These feeds reward early completion. A 21 to 34 second video that gets watched twice outperforms a 60-second video watched halfway. Build the short cut first: hook in the first two seconds, product by second eight, CTA by second twenty. Keep captions on, because a large share of viewers watch muted, and muted viewing changes the pacing calculation — text on screen reads faster than speech, so you can compress spoken lines and let the caption carry detail.

Paid social feed (square and vertical)

Feed ads tolerate 30 to 45 seconds and often need more explanation because the viewer did not choose to watch. Use roughly seven seconds for problem framing before introducing the product. Anything past 45 seconds needs a strong reason to exist, such as a demo.

YouTube pre-roll and skippable placements

Six seconds is the bumper standard; 15 to 30 seconds is the skippable sweet spot. Front-load the value proposition before the skip button appears. A useful structure is a 6-second bumper plus a 30-second extended cut built from the same scenes, so the two share assets and stay consistent.

Owned channels and landing pages

Here, 60 to 120 seconds is fine and often better, because the viewer arrived with intent. Add a chaptered structure: three or four scenes that each answer one question a buyer would ask.

A Repeatable Workflow From Brief to Export

Here is a sequence that consistently produces on-target runtimes.

1. Lock placement and target length. One primary placement, one primary runtime.
2. Choose the voice before writing. Voice sets the words-per-minute base rate for everything else.
3. Write to a word budget. Count words as you draft, not after.
4. Storyboard scenes with a per-scene time target. Assign a purpose to every scene; delete any scene with no purpose.
5. Generate one scene at a time. Check actual duration as you go. Fix drift at scene two, not at scene nine.
6. Review the assembled timeline. Look at total runtime including transitions, not just speech.
7. Adjust with the cheap levers. Trim a sentence, switch a transition, nudge voice speed.
8. Export the primary cut. Then duplicate the project and make the derived lengths.
9. Add captions and an end card hold. Confirm the last frame lingers long enough to read.
10. Log the final runtime. Keep a simple table of target versus actual so your estimates improve over time.

Step ten is the one most teams skip, and it is the one that makes future scripts accurate. After five or six videos you will know your own voice's actual rate, your own transition overhead, and your own typical overshoot.

Common Mistakes That Break Pacing

Writing to a page instead of a timeline

Prose that reads well often has subordinate clauses that add seconds without adding meaning. Read the script aloud with a stopwatch. If you stumble, the avatar will too.

Ignoring transition overhead

Ten scenes at half a second of transition each is five seconds of your 30-second budget. Account for it in planning.

Setting speed too high as a shortcut

Squeezing 20 percent more content into the same runtime by speeding up the voice hurts comprehension more than the extra content helps. Cut the content instead.

One length for every platform

A single 45-second master cropped for every channel usually underperforms a purpose-built 15-second vertical and a 45-second feed cut made from the same scenes.

No visual change during long speech

If a scene runs 12 seconds with no visual change, the viewer's attention drifts. Break it, or add a supporting visual element on screen.

Forgetting the muted viewer

If the message only lands in audio, the effective video length for a muted viewer is zero seconds. On-screen text is a pacing tool, not decoration.

Testing and Iterating on Length

Treat runtime as a variable to test, not a constant to defend. If you have two versions of a 30-second cut, one at 24 seconds and one at 38 seconds, run both with the same creative and compare completion rate, not just click-through rate.

Metrics worth watching alongside length:

• Hook retention: percentage still watching at three seconds.
• Completion rate: the share that reaches the end. Rising completion with falling views often means your cut is now too long for the placement.
• Average watch time: compare against your runtime to see how much is genuinely consumed.
• Cost per result: the only metric that translates length directly into budget efficiency.

A practical rule: if completion rate drops below roughly 25 percent on a short-form placement, test a cut that is 20 percent shorter before changing the creative idea. Often the concept is fine and the edit is simply long.

FAQ: Troubleshooting Video Length in Synthesia

Why is my render longer than the estimate?
Estimates usually do not fully account for punctuation pauses, scene transition overhead, or end-card holds. Generate a single scene, measure it, then extrapolate across your scene count.

Can I set an exact duration and force the video to fit?
Not directly. You control duration indirectly through script length, voice rate, pause length, transitions, and scene count. Aim to land within about one second of target, then use voice speed for the final nudge.

What is the fastest way to cut five seconds?
In order: delete one redundant sentence, convert two fades to hard cuts, reduce the end-card hold, then raise voice speed slightly. That sequence usually recovers five seconds without harming the message.

Does aspect ratio change runtime?
No, but it changes perceived pacing. Vertical formats cut faster in the viewer's mind, so identical scripts feel slower in 16:9 than in 9:16. Shorten vertical cuts slightly for the same energy.

Should every platform get its own export?
Every platform with a different viewing context should. Build a primary cut, then derive a 6-second bumper, a 15-second vertical, and a longer owned-channel version from the same scenes. Shared assets keep production fast and brand consistent.

How many scenes is too many for a 30-second video?
Past six or seven scenes, transitions eat your budget and the video feels like a slideshow. Consolidate to four or five purposeful scenes for 30 seconds.

How do I keep quality high while shortening?
Cut words before cutting ideas. If you must remove an idea, remove the least differentiated proof point and keep the hook, the core benefit, and the CTA intact. Those three elements carry the performance of the ad.

Length is not a technical detail you fix at the end. It is one of the earliest creative decisions you make, and in a tool like Synthesia it is controlled through writing discipline, scene planning, and a handful of audio and transition settings. Get those aligned before you generate, and pacing stops being a source of rework and becomes a repeatable part of how you ship marketing video.