Breeze TTS 2 is one of the most capable models Avocado works with: it takes direction, clones, designs and can be taught a new voice.
Speakers and direction
Choose one of its ten preset speakers and, if you like, add a direction for the take: “warm and unhurried”, “whispering”, “excited, a bit faster”. Guidance sets how strongly the direction is followed. Sound tags such as laughs and sighs can go right in the script.
Clone
Clone a voice from 1 to 30 seconds of clear speech with an exact transcript of what was said. Only clone voices you have permission to use; Avocado asks you to confirm before it makes a clone.
Train
Teach Breeze a voice from your own recordings: between one and thirty minutes of speech in 5 to 900 clips. The trained voice appears in your Voice Library with a Trained badge.
Speed
Measured on an M3 Max, Breeze generates speech in about 0.32 times its playing time (0.38 times with guidance), and playback starts after about 0.29 seconds, while the rest is still being made.
Licence
Breeze TTS 2’s licence is set by its maker. Read it on the Breeze TTS 2 model page before you publish anything you make.