An XTTS v2 alternative: what changed after Coqui, and what to use
XTTS v2 from Coqui is still one of the best-known open voice models: it clones a voice from a 6-second clip and speaks 17 languages. Two things make people look for an alternative: its weights are under the Coqui Public Model License, which allows only non-commercial use, and the company that could sell a commercial license closed at the turn of 2024.
The free SuperClonica engine runs the Supertonic 3 model in 31 languages on an ordinary processor, and its stock voices may be used commercially. This guide compares the two honestly and says when XTTS v2 is still the better pick.
XTTS v2 in short
XTTS v2 imitates a voice at generation time from a short sample, so any recording becomes a voice in seconds. It covers English, Spanish, French, German, Italian, Portuguese, Polish, Turkish, Russian, Dutch, Czech, Arabic, Chinese, Japanese, Hungarian, Korean and Hindi, and outputs 24 kHz audio. The code lives on in the community fork idiap/coqui-ai-TTS, installed as coqui-tts under the MPL 2.0 license, but that license covers the code, not the model: the CPML allows the weights and the audio they produce only where you receive no direct or indirect payment. It runs best on an NVIDIA graphics card.
Compare the sound free
On the page of any voice in the catalog, paste a line you voiced with XTTS v2, 30 to 100 characters, choose the language and press Voice it. In a few seconds you can compare, with no account.
What SuperClonica does differently
Voicing runs on an ordinary processor: in our tests a short phrase took 0.43 s on four threads of a Ryzen 7 5800X, and a graphics card only makes it faster. It speaks 31 languages, including Swedish, Greek, Ukrainian, Vietnamese and Indonesian, which XTTS v2 does not, but not Chinese yet. The ten stock voices, M1 to M5 and F1 to F5, may be used commercially under the OpenRAIL-M model license. Install it with python -m pip install superclonica-cpu, or call the same voices through the SuperClonica API without running anything yourself.
Your own voice
XTTS v2 clones from a sample every time it speaks. In SuperClonica a voice is trained once in the studio from a clean 5 to 10 second recording, on a computer with an NVIDIA graphics card, or ordered from an author who trains it. After that it is a small file that speaks offline on any processor. Clone only your own voice or one whose speaker has agreed; voices trained with consent may be used commercially, character voices only for fan and parody projects.
Which one to choose
Choose XTTS v2 for a hobby or research project that needs instant cloning from a sample, for Chinese, and when you have an NVIDIA graphics card. Choose SuperClonica for anything you are paid for, for languages XTTS v2 lacks, for ready character and narrator voices, and for simple deployment on ordinary servers. Nothing stops you from testing both on the same text first.