A Pocket TTS alternative: 31 languages, ready voices and a hosted API
Pocket TTS is a small open text to speech model from Kyutai: about 100 million parameters, real time on an ordinary laptop processor, and voice cloning from a short WAV recording. It speaks six languages: English, French, German, Portuguese, Italian and Spanish.
SuperClonica builds on Supertonic 3, a model of the same size class that speaks 31 languages, and adds a catalog of 190 voices made by authors, a free engine that runs offline and a hosted API. This guide compares them honestly and says when each one fits.
Pocket TTS in short
According to its model card, Pocket TTS has about 100M parameters, runs about 6 times faster than real time on a MacBook Air M4 processor using two cores, and gives its first audio chunk in about 200 ms. It clones a voice from a WAV file you pass to it; the card advises cleaning the sample first, because its quality carries over. The weights are published under CC-BY 4.0 and are gated: you accept the conditions on Hugging Face before downloading. Its rules forbid cloning a voice without explicit and lawful consent.
Where SuperClonica differs
Languages: Supertonic 3 speaks 31, including Russian, Japanese, Korean, Arabic, Hindi, Polish and Turkish, which Pocket TTS does not cover; neither speaks Chinese.
Voices: instead of cloning a clip at every request, SuperClonica voices are trained style files. You pick one of 190 catalog voices, listen to it and voice a test phrase online for free, or train your own voice from a recording in the studio.
Hosting: besides the free offline engine, the SuperClonica speech API runs the voices for you, streams audio from about 300 ms in our measurements and costs $5 a month for up to 216 hours of audio, with a free key for a minute a day.
Which one to choose
Choose Pocket TTS when you need instant cloning from a few seconds of audio in one of its six languages and you want to run everything yourself. Choose SuperClonica when you need more languages, ready-made character and narrator voices, or a hosted API with streaming and no server to maintain. Both run on a CPU, so you can try each on your own machine before deciding.