Kitten ML has announced the release of Kitten TTS V0.8, a new suite of open-source text-to-speech (TTS) models designed to be extremely compact and efficient.

Key Features

  • Small Size: The smallest model, Nano 14M, occupies only 25 MB.
  • Expressive Voices: Offers eight distinct voices (4 male and 4 female) with high expressiveness.
  • CPU Execution: Does not require a GPU, ideal for edge devices and resource-constrained systems.
  • Open Source License: Distributed under the Apache 2.0 license for free use.
  • On-Device Applications: Allows creating voice agents and local voice applications without the need for cloud APIs.

Available Models

  • Mini 80M
  • Micro 40M
  • Nano 14M

The Kitten TTS V0.8 models represent a significant step forward compared to previous versions, offering improved quality, expressiveness, and realism thanks to advanced training pipelines and larger datasets. The ability to operate on CPUs and the small size make them particularly suitable for applications where data sovereignty and low latency are crucial. For those evaluating on-premise deployments, there are trade-offs to consider, as highlighted by AI-RADAR's analytical frameworks on /llm-onpremise.