It states that training used more than 1,578 hours of synthetic speech generated by VoxCPM2 and no real speakers. The model is listed under the Apache-2.0 license, including commercial use. These are claims from the model card; the supplied source does not independently verify the training corpus or report an external evaluation.
Raw SignalOct 11, 20261 min read1 source
Trained on 1,578 hours of synthetic speech generated by VoxCPM2 , no real speakers used.
In one minute
OpenBMB’s VoxCPM2 model card describes a 2-billion-parameter text-to-speech model supporting 30 languages and 48 kHz audio output, with voice design and voice-cloning features.