Regarding the rule for AI generation, what is the verdict for a voice synthesizer like VoiceVox? It is mainly built on voice samples provided by real paid talents like the Vocaloids, and the wiki has the provider name for every model. However, it also makes use of deep learning (presumably to handle tricky conjugations and adopt a more natural tone in sentences?). As far as I know the training data is the voicebank the company builds themselves, but I can’t be certain they use no pretrained models with ethical concerns.