Radar ·
Open source TTS models at half the size of Kokoro, with voice cloning
An author announces open source text-to-speech models about half the size of Kokoro (82M parameters), reporting higher scores for naturalness and accuracy and support for voice cloning. No link is provided, so the results have not been verified.
According to the post dated October 20, 2026, the author announces open source text-to-speech models about half the size of Kokoro, which has 82 million parameters. The text states that the new models score higher on naturalness and accuracy and support voice cloning. The source gives no benchmarks, metrics, licenses or links, so these claims belong to the author and are not verified data.
The central idea is that smaller models run locally (self-hosted) could reduce the cost of speech synthesis. Anyone evaluating adoption should read the original publication, compare the results against their own test set, and check the license before any use. Voice cloning treats an identifiable voice as personal data: without documented consent, real recordings of people should not be used to imitate their voices. Rota Nacional does not present this model as its own capability, and voice remains in preview on the platform.
Original source: Modelos TTS open source com metade do tamanho do Kokoro e clonagem de voz