Kyutai TTS alternative: more languages, your own voices, no GPU
Kyutai TTS is an open streaming text-to-speech model from the French lab Kyutai: it starts speaking after the first few words of text, which makes it attractive for voice assistants. It speaks English and French, and its voices come as pre-computed embeddings.
This guide sums up what the model card says, when Kyutai TTS is the right pick, and when SuperClonica, built on the small Supertonic 3 model, fits better.
Kyutai TTS in short
Despite the 1.6b in its name, the model card says it has 1.8 billion parameters: a 1 billion backbone and a 600 million depth transformer. It speaks English and French and streams: audio starts as soon as the first words arrive. The weights are released under CC-BY 4.0, it runs with Kyutai’s moshi Python package, and the example code uses a CUDA GPU.
Voices and cloning
Kyutai chose to restrict voice cloning to pre-computed voice embeddings, published in a separate tts-voices repository, so you pick from the voices they provide rather than cloning any recording. The model card also notes there is no watermarking.
Where SuperClonica differs
SuperClonica runs Supertonic 3, a much smaller model, on an ordinary processor with no GPU, and every voice speaks 31 languages, from English and French to Russian, Japanese and Korean. The catalog has narrators and character voices, ten stock voices are free, and the studio trains a voice from your own clean 5 to 10 second recording. For apps, the hosted Speech API streams audio too, and our streaming API guide shows how.
Which one to choose
Pick Kyutai TTS if you build an English or French assistant on your own GPU and want its streaming design and open weights. Pick SuperClonica if you need more languages, a laptop or server without a GPU, character voices or your own voice. Kyutai also makes the smaller Pocket TTS, which our Pocket TTS guide compares. Try any SuperClonica voice free on its page with 30 to 100 characters, or install the engine with python -m pip install superclonica-cpu.