Voices
54 presets
Languages
9, including US and UK English, Spanish, French, Hindi, Italian, Japanese, Brazilian Portuguese and Mandarin
First sound
0.16 to 0.21 s on an M3 Max
42-second paragraph
Ready in 0.7 to 0.9 s on an M3 Max
Memory
About 2 GB
Voice training
Not available
Made by
hexgrad
Mixing two Kokoro voices, Heart at 60 percent and Michael at 40 percent

Kokoro is the quick one. It’s small enough to start almost instantly and fast enough that a long paragraph is ready before you’ve finished reading it back.

Voices

Kokoro has 54 preset voices. Avocado groups them by language in the voice picker, best-rated first, with a note such as “Female · Best”, and you can audition any of them in its own language from the Voice Library.

Mix a voice

On Create Voice, choose Mix, pick two to four of Kokoro’s voices and set how much of each goes in. The first voice sets the language; the others lend their character. Audition the mix, save it, and it appears under Your voices.

Pace

A Pace control slows a voice to half speed or speeds it up to double. The pace is saved with each take.

Speed

Measured on an M3 Max with the model loaded, the first sound of a one-sentence take plays in 0.16 to 0.21 seconds, and a 42-second paragraph is finished in 0.7 to 0.9 seconds.

Licence

Kokoro’s licence is set by its maker. Read it on the Kokoro model page before you publish anything you make.

Questions

Can I make my own voice with Kokoro?

You can mix two to four of its voices into a new one. Avocado can also fit the closest Kokoro voice to a recording of you, as an experiment; the result is an approximation built from Kokoro's own voices, not a copy of yours. For a real copy of a voice, use a model that clones, such as Fish Audio S2 Pro or Qwen3-TTS.

Is Kokoro included with Avocado?

No. Avocado works with models you add to your Mac yourself. Get Kokoro from its maker's page, and check its licence there.

Avocado is almost here

Free for Apple silicon Macs running macOS 15 or later. Coming soon.

Coming soon for Mac