Built on shared work

Credits & references

Kenkui brings together open software, research, and recorded voices. These are the works behind the project and the samples you hear here.

Pocket-TTS · Kyutai
The CPU speech-synthesis engine used by Kenkui.
Kyutai voice prompts
The source prompts used to compile the Kenkui voice catalog.
CSTR VCTK Corpus 0.92
Junichi Yamagishi, Christophe Veaux and Kirsten MacDonald, University of Edinburgh. Source of the 47 voices auditioned here; CC BY 4.0. Prompts were compiled into speaker embeddings and used to synthesize these recordings.
EARS dataset
An additional source in the wider Kenkui catalog, under CC BY-NC 4.0. EARS voices are not included in this site’s auditions.
Pride and Prejudice · Jane Austen
Our example scene is an excerpt from chapter 1. Narrator: Vivienne (VCTK p231); Mrs Bennet: Beatrix (p233); Mr Bennet: Rex (p243). Read the original at Project Gutenberg.
spaCy
Local character discovery and language processing.
LiteLLM
The provider interface for language-model-based dialogue attribution.
FFmpeg
Audio encoding and packaging for the finished M4B audiobook.

Voice and model terms are separate from the software licenses. The catalog marks built-in voices as not cleared for commercial use by default. See the voice rights guide and the original CC BY 4.0 and CC BY-NC 4.0 terms.