Open-source Qwen3-TTS 1.7B CustomVoice voice cloning model
A post presents an open-source 1.7B text-to-speech model for voice cloning, with controls for emotion, age, tone, accent and multilingual support. Its realism and latency claims remain unverified.
According to the post, Qwen3-TTS 1.7B CustomVoice is an open-source text-to-speech model with 1.7 billion parameters, aimed at voice cloning. It offers controls for emotion, age, tone and accent, plus multilingual support. The primary source is the model card published on the model platform, which records the official details. The author's claims about realism and latency have not been verified and should be treated as claims until independent testing.
For teams evaluating self-hosted TTS, the verification path is to read the license, the supported languages and the list of controls on the model card, then test with your own authorized samples. Cloning a real person's voice requires consent and raises fraud and identification risks, so voice is sensitive personal data for the organization. Rota Nacional does not present this model as its own capability; the platform applies the same personal data barrier to audio transcription and document extraction, and voice features are in preview.