Voice

Writing a script

Scripts are spoken exactly as written, in the language they are written in, by one of 20 stock voices (ElevenLabs Eleven v3, 70+ languages including Hebrew) or by your cloned voice. Square-bracket tags direct the delivery and timing; they are never read aloud. In the studio they appear as badges above the editor: click or drag one into the text.

Cue badges

Put a cue inline where the delivery should change:

[laughs][chuckles][whispers][excited][cheerful][calm][serious][sad][angry][surprised][curious][sarcastic][nervous][sighs][gasps][shouts][clears throat][pause]

Welcome to Frenkel Studio. [excited] Today we are making something special.
[whispers] And nobody has to learn a single setting. [laughs]

Timing tags

Tag	Effect
`[at:5]` · `[at:1:05]`	Starts the next line at that second of the clip (minutes:seconds allowed). Useful for voice-overs that must hit a moment in the video.
`[wait:1.5]`	Leaves that many seconds of silence before continuing.
`[break:0.8]`	A short pause inside a sentence.

Dialogue

Up to four speakers. Put [voice:2] before the lines of speaker 2 and map speakers to voices; speaker 1 is the main voice.

[voice:1] Did you see the new studio?
[voice:2] [surprised] Already? Show me.
[voice:1] [laughs] Sit down, this takes a minute.

In the API: "voice": {"voice": "Sarah", "speakers": [{"key": "2", "voice": "Roger"}]}. For lip sync of a dialogue, each speaker can carry a face point (fractions of the frame) so the right mouth moves.

Voices

Stock voices: Sarah, Aria, Laura, Charlotte, Alice and fifteen more, each with a one-line description in the picker. A cloned voice is trained from 10 seconds to 5 minutes of clean speech and speaks 40 languages; use it as clone:<id> in the API. Delivery (stability, from expressive to steady) and speed are simple controls; the script language is detected unless you set it.

Voice-over, lip sync and dubbing

Voice-over (video_voiceover): lay a script or an existing clip on a video. Keep the clip's own sound underneath (at 25% by default) or replace it; when the voice outlasts the clip, hold the last frame, slow the clip, or cut the voice.
Lip sync (video_lipsync, or voice.lipsync on a voice-over or dub): redraws the mouth to match the new speech. cloud uses sync-3 (best, priced per second of video); local uses InfiniteTalk on our GPU (cheap draft tier, 480p, about 9 seconds, re-animates motion from the first frame). Needs a clearly visible face.
Dubbing (video_dub): translate and re-voice a video into one of 22 languages. Languages ElevenLabs Dubbing supports keep the speaker's own voice; the others, Hebrew included, are transcribed, translated and spoken by the voice you choose. Set the source language when the clip is very short.

HebrewEnglishArabicRussianFrenchSpanishGermanItalianPortugueseDutchPolishTurkishUkrainianGreekRomanianCzechSwedishHindiChineseJapaneseKoreanIndonesian

Limits

Scripts up to 2,500 characters per clip; up to four speakers. Speech is priced by length (about 3 credits per 1,000 characters); voice cloning is a one-time fee; lip sync and dubbing are priced by the length of the video. See Credits & pricing.