Black Friday sale· 50% off first month
A studio microphone with a pop filter and a glowing sound wave in a dark recording booth
Audio

Official BytePlus partnerBytePlus, a ByteDance company

How to add a voiceover to an AI video

Add a voiceover to a Seedance video: size the script to the clip, voice it in Text to Speech at 2 credits per 1,000 characters and edit it in.

KreatorFlow7 min read

Short answer

Generate the clip with Seedance 2.0 / Seedance 2.5 in the video studio, write a script that fits its length, turn it into a voiceover in Text to Speech (2 credits per 1,000 characters) and lay it under the clip in your editor, or attach it as an audio reference. Native Seedance 2.x audio adds no extra credits; the voiceover is the scripted part.

Updated:

Step by step

  1. 1

    Decide the length

    Seedance 2.0 clips run from 4 to 15 seconds. A natural read is about two to three words per second, so a 10-second clip holds roughly 25 words and a 15-second clip about 38. Pick the clip length first; the script has to fit it, not the other way round.

  2. 2

    Generate the video

    Make the clip in the video studio. Five seconds at 720p with Seedance 2.0 costs 17 credits. With native audio on, the clip brings its own sound at no extra credits. If a voice will carry the message, ask for a calm shot: one subject, one clear action and no on-screen text the voice has to compete with.

  3. 3

    Write the voiceover

    Short sentences, one idea each, the product early and the call to action last. Write for the ear: contractions, plain words and numbers spelled the way they are said.

  4. 4

    Read it against the clip

    Play the clip and read the script out loud with a timer running. If you run over, cut words rather than speeding up. This one read catches most timing problems before you spend anything.

  5. 5

    Generate the voice

    In Text to Speech, pick one of 45 voices in 14 languages and set the speech rate between 0.5x and 2x so it fits. A 300-character script costs 2 credits. Add a style instruction of up to 300 characters, such as "warm and confident, like a product launch", to steer the delivery.

  6. 6

    Fine-tune the delivery

    Open Advanced to adjust tone and pitch (0.5x to 2x) and volume. Voices that support it also offer an Emotion setting with an intensity slider. Change one setting at a time so you know what made the difference.

  7. 7

    Put picture and voice together

    Download the voiceover and place it under the clip in your editor, or attach it in the video studio as one of up to 3 audio references and describe in the prompt how to use it.

Example scripts

Paste them into Text to Speech and swap in your own product.

  • 10-second ad

    Tired of tangled cables? This charger powers your phone, watch and earbuds at once. Try it today.

  • Explainer opener

    In the next minute you will see how to go from an idea to a finished video. Let's start with the script.

  • App walkthrough

    Open the app, tap New, and type what you want to see. That's it. Your first draft is ready before your coffee is.

  • Story hook

    Nobody on this street had seen snow in forty years. Then, one Tuesday morning, everything changed.

Native audio and a voiceover: what each is for

Seedance can generate sound with the picture: ambience, effects and the feel of the scene. On Seedance 2.x that native audio adds no credits.

The voiceover is what you control word for word: the exact script, the voice and the pace. Most videos use both: native audio for the room and the action, a voiceover for the message.

If the clip has people talking on screen, keep the voiceover for the moments where nobody speaks, or turn native audio off for that clip so two voices never talk over each other.

How long should the script be?

Count words, not characters. At a relaxed two and a half words per second, these are good targets for a single clip. Leave the first half second and the last second free so the voice does not start or end on a cut.

  • 4 s clip: about 10 words, one sentence
  • 5 s clip: about 13 words, a hook or a tagline
  • 10 s clip: about 25 words, a problem and a promise
  • 15 s clip: about 38 words, a short ad with a call to action
  • Longer videos: write one script for the whole edit and generate it in parts, one per scene

A script structure that works for short ads

Most short ads and explainers follow the same four beats: a hook that names the problem, the product as the answer, one reason to believe it, and one call to action. Keep each beat to a sentence. If you need a second call to action, you need a second video.

Here is a full script for a clip at the longest Seedance 2.0 length, written to that structure. Read it aloud against the timer before you generate it.

Four-beat ad script

Your coffee goes cold before you finish it. Meet the mug that keeps it hot for three hours. It survives the school run, the commute and the meeting that ran long. Tap the link and get yours today.

Worked example: one product clip, start to finish

Say you sell a desk lamp. You generate a 5-second 720p shot of the lamp switching on in a dark room with Seedance 2.0 for 17 credits. Native audio adds the click and the room tone at no extra cost.

You write an 11-word line, "Late nights, sorted. One touch, warm light, zero glare. Shop now.", which fits the five seconds with room to breathe. It is well under 1,000 characters, so it costs one block: 2 credits. Generate two takes at slightly different speeds for 4 credits and keep the one that lands on the cut.

The finished clip with a voiceover costs 21 credits in total, and the studios show the cost of each step before you generate.

Getting the delivery right

The same words can sound like an ad, a documentary or a friend's recommendation. Three settings decide which.

  • Voice: choose by the audience, not by your own taste. Narration voices suit explainers; brighter voices suit social ads.
  • Style instruction: describe the read in a few words, such as "calm, close to the mic, like a late-night radio host".
  • Speech rate: slow down slightly for prices, steps and names; speed up a touch for energetic hooks.
  • Emotion: on voices that offer it, keep the intensity low at first. Strong emotion on every line sounds forced.

Tips for a voice that fits the cut

Small habits that save a re-edit.

  • Write numbers and prices the way they should be said.
  • Use punctuation for pauses: a comma for a breath, a full stop for a longer beat.
  • Generate two or three takes at different speeds and keep the one that fits.
  • Set the voice volume so it sits clearly above the music.
  • Start the voice slightly after the cut, never exactly on it.
  • End on the call to action, then let the picture hold for a moment of silence.

Mixing voice, music and native audio

In your editor, treat the voice as the loudest layer. Native ambience sits underneath it and music sits underneath both. When the voice speaks, lower the music; when it stops, the music can come back up.

You can do this in the free KreatorFlow video editor: put the clip on the timeline, add the voiceover from the Audio tab with "Upload audio", and adjust each track's volume.

Common mistakes

These are the reasons most voiceovers get redone.

  • Writing the script first and then trying to squeeze it into a short clip
  • Speeding up the voice to make a long script fit instead of cutting words
  • Reading out on-screen text word for word, so picture and voice say the same thing twice
  • Two calls to action in one clip
  • Leaving native dialogue on under a voiceover, so two voices compete
  • Abbreviations and symbols the voice reads literally, such as "approx." or "&"

Voiceovers in other languages

Text to Speech has preset voices in 14 languages, and Audio Generation speaks 24 languages from a written prompt. For a second market, write the script in that language rather than translating it word for word: idioms, sentence length and the call to action all change.

Many languages need more words than English to say the same thing. Check the timing again for every language and trim the script, not the speech rate.

Frequently asked questions

Is the voiceover included in the video price?

No. It is a Text to Speech clip at 2 credits per 1,000 characters, with a minimum of 2 credits. Native Seedance 2.x audio adds no extra credits.

Can I make the voiceover in another language?

Yes. 14 languages have preset voices, and Audio Generation speaks 24 languages from a prompt.

Can I use my own voice?

Yes. Clone your voice from 10 to 30 seconds of clean speech for 60 credits, with your consent recorded.

How many words fit in a short AI video?

About two and a half words per second: roughly 13 words for 5 seconds and 38 words for a 15-second clip.

How long can one voiceover be?

One Text to Speech clip takes up to 5,000 characters, enough for several minutes of narration. For a long video, generating one clip per scene makes re-edits cheaper.

Why does my voiceover sound rushed?

The script is too long for the clip. Cut words until a read-through at a normal pace fits, then set the speech rate back near the default.

Can I change how the voice sounds without changing the voice?

Yes. Use the style instruction for the overall read, the speech rate for pace, and Advanced for tone, pitch, volume and, on voices that support it, emotion.

Try it in the studio

New accounts get 110 free credits (10 on sign-up + 100 for linking Discord) for images. Video starts on the Creator plan, and the cost is shown before every generation.

More guides