Glossary
Text-to-speech (TTS)
Technology that converts written text into spoken audio. Modern neural TTS voices sound nearly human and power most talking-avatar tools today.
Text-to-speech (TTS) maps a string of text and a target voice to a waveform. Early TTS sounded robotic; modern systems use deep neural networks trained on hundreds of hours of human speech and produce natural prosody, breaths, and emotion.
In VlogMe you can pick a TTS voice, clone a voice with its owner's consent, or upload a recording — the talking-avatar workflow then lip-syncs the portrait to whichever audio you choose.
How neural text-to-speech works
A neural TTS system first works out how the text should sound: how each word is pronounced, which syllables carry the stress, where the voice rises, where it pauses. It then generates the waveform in the chosen voice. Multilingual models can read many languages in one voice, and expressive models react to punctuation, emphasis and the meaning of the sentence.
Three kinds of voice are common:
- Library voices, recorded by voice actors who licensed them for synthesis.
- Cloned voices, built from a short sample of one person's speech, with that person's consent.
- Designed voices, generated from a written description instead of a recording.
Text-to-speech in VlogMe
Audio Studio voices a text of up to 5,000 characters in one run, on the ElevenLabs v3 and Multilingual v2 models. Pick a library voice, or clone one from a clean recording of up to 8 MB whose speaker agreed to it. Billing counts each started block of 1,000 characters, and you see the price before you confirm. Because the models are multilingual, a single voice can switch languages and still sound like itself.
To give the voice a face, use it in Voice to avatar video: a portrait speaks the track with lip sync, for as long as the audio runs.
Limits and risks
- Names, abbreviations and numbers can be read the wrong way: spell them out, or write them the way they sound.
- Long texts sound more natural when they are generated and checked paragraph by paragraph.
- A cloned voice may be used only for what its owner agreed to, including the languages and where the result appears (Terms, section 6).
Questions
What is the difference between TTS and voice cloning?
TTS reads text aloud in a voice. Voice cloning creates that voice from a sample of a real person; the clone is then used through TTS like any library voice.
Which languages can a voice speak?
In VlogMe, the languages the ElevenLabs models support, in the same voice.
Can I edit the speech after it is generated?
Change the text and generate the passage again; each run shows its price first. Recordings in Lab can also be cleaned up or transcribed in Audio Studio.
