25launches

The first launch of Kitten-TTS 2.

Kitten-TTS 2 page

Kitten-TTS 2

Launching today

Open-source text-to-speech that runs on CPU, from 4 MB

Hunted by IlyasHunter
Visit kittenml.com (opens in a new tab)

Kitten-TTS 2 is KittenML's family of text-to-speech models that run on a CPU. The new Pico (7M parameters, 4 MB per language in int4, 8 languages), Nano (17M) and Mini (82M, 16 languages, 118 voices) are Apache-2.0 ONNX models, installed with pip. The larger 1.7B model adds voice cloning from a few seconds of audio, emotion tags and 20 languages, but its license is free only below $1M in revenue and funding. KittenML also sells a hosted API.

Meet the team

Hunted by Ilyas. The makers are not on 25launches yet.

H (Hunter)
  • Kitten-TTS 2 screenshot 1
  • Kitten-TTS 2 screenshot 2
  • Kitten-TTS 2 screenshot 3

Discussion 1

Ask questions, give feedback. Comments don't count in the ranking.

Sign in to join the discussion.

  • IlyasHunter Pinned

    Good text-to-speech usually means a GPU or a per-character cloud bill. The smallest Kitten-TTS 2 model is 4 MB per language, and KittenML says it runs faster than real time on a single phone CPU thread. Check the license before shipping: the small models are Apache-2.0, but the 1.7B cloning model is free only for small companies. The GitHub repo is at almost 17,000 stars.