ElevenLabs — text to speech and video dubbing

What ElevenLabs is

ElevenLabs is a company specialising in artificial intelligence for voices and voice-over. Its technology synthesises natural-sounding speech and can recreate the characteristic traits of a human voice — tone, intonation and timbre — in many different languages. Thanks to the quality of those voices, ElevenLabs has become one of the leading names in AI voice-over and dubbing.

In genkiki.com ElevenLabs is available through the Dubbing mode — a tool that takes a finished video with speech and translates and voices it in another language, keeping a voice close to the original speaker.

Versions and capabilities in generiram

ElevenLabs comes in two modes: video dubbing and text to speech.

When dubbing, the speech from the source video is translated and voiced again in the language you choose. The voice is recreated so that it sounds close to the original speaker — that is, the dub does not sound like a stranger's voice, but like a translation spoken by the same person. The lips are not reshaped to the new language, which makes the processing faster and cheaper than full dubbing with lip-sync.

• Up to 29 languages for the voice-over.
• Video up to about 10 minutes long.
• Automatic detection of the source language — you don't have to set it by hand.
• The option to state how many speakers there are in the video, for a more accurate result.
• Voice cloning, so the translation sounds close to the original speaker.

What it is good for

ElevenLabs Dubbing is most useful when you want to make a video understandable in another language fast, without reshooting anything:

• Quick voice dubbing of videos for a new audience and a foreign market.
• Translating training clips, presentations and webinars.
• Voicing marketing and product videos in several languages.
• Situations where the lips are not in close-up — there the missing lip adjustment is barely noticeable, and you gain from the faster, cheaper processing.

ElevenLabs v3 — text to speech

ElevenLabs v3 is the voice model in the Audio category of the AI Studio: you write text, pick a voice and get a finished audio file, as if a professional narrator had read it. There are 21 voices — male and female, with different characters — and each one has a sample you can listen to before choosing.

• Text is limited to 5000 characters per generation — the counter under the field shows where you are.
• You are charged per 1000 characters started — 1 to 1000 characters cost the same as 1000.
• You can drop in cues in square brackets that are not read out but acted out — write them in your own language and the system translates them for the model: [laughs], [whispers], [sighs], [excited], [sarcastic], [crying], [singing], [pause] and more.
• Reading style: Creative — more expressive and emotional; Natural — balanced; Stable — even and predictable (for narrators and lessons).
• The language is usually detected from the text itself — set it by hand only if the pronunciation comes out wrong. It supports dozens of languages.
• The text is read word for word — nothing is translated or shortened. The “Numbers and abbreviations” setting decides whether “€25” is spoken as “twenty-five euros”.

Open the AI Studio, choose the Text → Speech mode in the Audio category, write your text and press Generate — seconds later the audio is ready to download.

How to get started

Open the AI Studio at genkiki.com, choose the Video → Video mode and the ElevenLabs Dubbing variant. Upload the video, pick the target language and, if needed, state the number of speakers. The source language is detected automatically. Start the processing and you will get the same video, voiced in the new language with a voice close to the original.