DocumentationUsing the dub
Speakers and voices
How myDubber tells speakers apart and gives each one a voice of their own.
myDubber doesn’t use one narrator for everybody. It tells the people in a video apart and gives each one a voice cloned from their own.
How it works
As the video plays, myDubber:
- separates the voices from the music and sounds;
- recognises what is said;
- works out who is speaking, and whether this person has spoken before;
- translates, knowing the speaker’s gender, so the grammar fits;
- speaks the translation in a voice modelled on that speaker;
- mixes it back under the original music and sounds, in sync with the picture.
Where the dub doesn’t replace the original voice, you still hear the original.
The Speakers list
The popup’s Speakers section lists everyone myDubber has heard so far, with a counter. For each speaker it shows:
- Label
- Meaning
- The name myDubber gives this speaker
- Language code
- Meaning
- The language this person speaks, once clear
- male, female or ?
- Meaning
- The voice’s gender, once clear
- Seconds
- Meaning
- How much of their speech myDubber has heard
Until someone talks, the section says Speakers appear once someone talks.
Gender shows ? at first. myDubber decides only after it has heard enough of a person’s speech.
Choosing voices
There is no voice picker. Each speaker always gets a voice modelled on their own, so two people in a conversation sound different, and a woman is voiced as a woman.
About cloned voices
The voice you hear is synthetic. It is modelled on the speaker so the dub sounds like them, but it is not their real voice and it says translated words, not theirs. It is made on the fly for your own viewing.
Please don’t use myDubber to make content that presents a synthetic voice as something a real person said.