Contents
A video often looks finished before it sounds finished. Voice and captions are added in the same project, with no separate service.
Voiceover is created in the Voiceover mode of Generation: from text, or from an existing recording with the voice replaced. The finished clip goes on the Voiceover track in the Editor, and captions come from the voiceover text or a video transcription.
Add voice once the visual rhythm is roughly clear but the video is still easy to change: a voiceover often changes scene length and where text fits in the frame.
What you need
- A project with a rough visual sequence or at least one scene.
Step by step
- 1
Build the visual rhythm
Arrange the main scenes before finalizing narration.
- 2
Draft in Eleven v4 Turbo
Write the text with tags and listen to the draft: pace, pauses and emotions. Turbo costs half as much as v4.
- 3
Voice the final take in Eleven v4
Regenerate the approved text in v4 with the same tags. The voiceover goes on the Voiceover track in the Editor.
- 4
Build captions
Pick a style and position, and check that the text is readable.
- 5
Review with and without sound
Scenes, voice, text and pace should work together, and captions should carry the message when muted.
Where the voice comes from
Voiceover from text
Write the text, pick a model and a Voice from the base set or your own. Every voice can be previewed before generation.
Voice replacement
Eleven v2 replaces the voice in an existing recording and keeps the speech, pace and intonation. Pick the Voiceover source: Project audio/video, Audio or video file, or Microphone, up to 5 minutes.
Your own voice
Library → Voices → My voices → New voice: add 1–2 minutes of clean speech. Voice training requires an active plan, and the finished voice appears in the Voice picker.
Which model to choose
- Eleven v4The main model for final voiceover: follows tags more closely and takes the context of the whole text into account. Up to 2,000 characters recommended; up to 10,000 accepted with a warning. The price is 8 tokens per 1,000 characters.
- Eleven v4 TurboThe same tags and settings at half the price: 4 tokens per 1,000 characters. Handy for drafts to check the text, pace and tag placement.
- Eleven v3Up to 4,000 characters recommended; up to 5,000 accepted with a warning. Ready-made Delivery presets: Natural, Calm, Expressive, Ad.
How to set delivery with tags
Tone is set with square-bracket tags right in the text. Type “[” in the voiceover field and Givon suggests 12 tags: [whispers], [excited], [laughs], [sighs], [sad], [angry], [curious], [sarcastic], [crying], [shouting], [pause] and [exhales].
Your own tags work too: any text of up to 40 characters in square brackets goes to the model. Tags count toward the request's characters.
How to pick a voice, place tags and pauses, write numbers and which templates to use for ads, explainers and characters is in the voiceover script guide.
Tag before the phrase
A tag changes the part of the text that follows it. When the emotion changes, add a new tag.
Stacked tags
[sighs] [sad] means a sigh first, then a sad line. Eleven v4 plays tags in order.
Pauses and emphasis
An ellipsis adds a pause and weight to a phrase, and a word in CAPITALS is stressed.
A description instead of an emotion
If a tag came out as a sound, describe the delivery: instead of [crying], write [trembling, tearful voice].
Sounds in the text
Eleven v4 also understands sound tags such as [applause] or [door slams]. [sings] and accents like [strong French accent] do not work on every voice.
A voice that fits the delivery
A whisper works best with a soft voice, a shout with an energetic one. Preview a few voices on the same text.
[sighs] Here we go again... [sarcastic] Sure, that was a VERY good idea. [exhales] Fine. [curious] So what is it this time?
Voice settings
Instead of Delivery, Eleven v4 and v4 Turbo have two percentage settings. Besides the preset values, each has a Custom value option.
- Expressiveness
- 50% by default. Higher is more expressive but varies more between takes; lower is steadier and more predictable.
- Similarity
- 75% by default. How closely the voiceover matches the selected voice; a high value also carries flaws of the source recording.
Pronunciation
Words a voice gets wrong are fixed in Library → Voices → Glossary: fill in “Word or phrase” and “Pronounce as”, then check it with Preview. A rule applies to every voiceover generated from text.
How to build captions
Captions are built in the Subtitles panel. A voiceover from text already carries its text and word timings; for a video with sound, press Transcribe, and the transcription cost is shown on the button before it runs.
Each cue can be edited, split, locked or deleted, and the whole track gets a style (Broadcast, Clean pop or Soft karaoke) and one of nine on-screen positions.
FAQ
Which model should I use for voiceover?
Eleven v4 for the final voiceover, Eleven v4 Turbo for drafts, and Eleven v3 when you want Delivery presets.
How do I make a voiceover more emotional?
Put tags before the phrases, such as [excited] or [whispers], and raise Expressiveness. Tags can be stacked: [sighs] [sad].
Can I use my own voice?
Yes. Create a voice in Library → Voices → New voice from 1–2 minutes of clean speech, or record a line with the microphone and replace the voice in it with Eleven v2.
How much does a voiceover cost?
Eleven v4 and v3 cost 8 tokens per 1,000 characters; Eleven v4 Turbo costs 4. The price rounds up to a whole token, with a 1-token minimum per generation. For example, 2,500 characters cost 20 tokens in v4/v3 or 10 in Turbo.
