Higgs Audio v3 from Boson AI is built for expressive speech. It reads the mood from the text and from the tags you add.
Tags
The script’s tag button lists all 43 of Higgs’ tags in groups: emotion, style, pace and pitch, pauses and sounds such as laughter. Pick one and Avocado puts it in the right place: emotions at the start of the sentence, pauses and sounds where your cursor is.
Voices
There are no preset voices. Without a cloned voice, Higgs makes up a new voice for each seed, and the same seed always gives the same voice. Clone a voice from up to eight clips to use one consistently; any voice you’ve cloned for another model works too. Only clone voices you have permission to use.
Long scripts
Send a whole script at once. Higgs keeps one voice across it, and if a part stalls Avocado re-tries it quietly rather than failing the take.
Playback
Takes play as they’re made, so you start hearing the result before the whole script is finished.
Licence
Higgs Audio v3’s licence is set by Boson AI. Read it on the Higgs Audio v3 model page before you publish anything you make.