| Why narration | Voiceover doubles as script discipline: if it is awkward to say, the shot is wrong |
| The 2026 default | Voiceover over B-roll — the most robust pattern (no lip-sync dependency) |
| Captions | Essential: most social video is watched muted; captions lift completion rates |
| In Plugsky Studio | Generate voiceover from script, pick the take, mux it into the assembled video |
| Pacing rule | ≈2.5 words per second; a 30s video = a 75-word script |
| Accessibility | Captions are also the accessible default — treat them as required, not optional |
Why voiceover is the script editor
Writing narration first exposes weak video ideas faster than anything else. If a shot cannot be described in a natural spoken sentence, it probably does not belong. The professional 2026 order is: script → storyboard → generate → voice → captions — the script constrains the storyboard, and the storyboard constrains the prompts. Doing it in reverse (generate first, narrate later) produces videos that sound like captions over clips, because that is what they are.
Choosing a voice
| Use case | Voice direction | Pacing |
|---|---|---|
| Product ad | Confident, warm, medium pace, slight energy lift on the CTA | ≈2.6 words/s |
| Tutorial/explainer | Clear, friendly, even — prioritize comprehension over energy | ≈2.2 words/s |
| UGC-style ad | Conversational, close-mic feel, natural imperfections | ≈2.8 words/s |
| Brand/launch film | Measured, cinematic, lower pitch | ≈2.0 words/s |
| Multilingual campaigns | Same voice family across languages for brand consistency | per language norms |
In Plugsky Studio the Voice & Audio view generates narration from your script, keeps multiple takes side by side, and muxes the chosen take into the assembled video. One rule from branding: pick one voice per character/brand and never mix — voice drift reads as inconsistency just like face drift.
Writing for the ear (not the eye)
- Short sentences. One idea per sentence; the ear cannot re-read.
- Numbers rounded. "About three times faster" beats "3.17x" in speech.
- No visual references. "As you can see" fails in narration — say what the viewer sees.
- CTA last and single. One action per video: "Start your free trial today" — not three asks.
- Read it aloud once. Every stumble is an edit you just found for free.
From takes to final audio
Music note: keep background music under the narration (roughly −18dB or "clearly audible words at low volume"), and duck it during the CTA. Most stock music fails ads because it competes with the voice — restraint sells.
Captions done right
- Short lines: 2–4 words per caption card, synced to speech rhythm.
- High contrast: white text with a soft shadow or dark backing; avoid thin fonts.
- Safe zones: keep captions out of the bottom 15% (UI overlays) and top 10% (profile elements).
- Punctuate for emphasis: a caption card is a beat — use line breaks where the voice pauses.
- Proofread twice. Auto-captions mishear brand names first — always fix names and numbers.
Narration and captions built into the studio
Plugsky AI Video Studio generates voiceover takes, muxes the best one, and times captions to your script — all before you export.
Open AI Video Studio → See pricingFrequently asked questions
Do AI videos need voiceover?
Not always — but narration massively improves retention and clarity for ads, explainers and social content. Product B-roll can work music-only; anything with a message benefits from voice.
Is voiceover the same as lip-sync?
No. Voiceover narrates over B-roll and product shots — the most robust pattern in 2026. Lip-sync (an on-camera character speaking) is a different, harder capability; not every studio supports it yet.
How long should my script be?
About 2.5 words per second of video: a 30-second video needs a ~75-word script. If your script is 150 words, the video is 60 seconds — decide which one changes.
Can I use AI voiceover commercially?
Yes with generated voices you are licensed to use — check the voice provider's terms. Avoid cloning real people's voices without written permission.
How do I add captions to AI video?
Generate captions from your narration script (more accurate than speech-to-text), keep 2–4 words per card, and burn them in for social. Plugsky Studio times captions to the script inside the studio.
What music should I use under narration?
Instrumental, low-energy under the voice (≈−18dB), with a lift at the CTA. Always verify licensing for commercial use.
Plugsky (2026). “AI Video with Voiceover & Captions (2026 Guide)”. Plugsky. Available at: https://plugsky.com/blog/ai-video-with-voiceover-2026 (last updated 2026-10-16).