Kotori (Rewrite) — TTS Fine-tune vs Zero-shot Comparison
5 model variants · each shown FT (fine-tuned on 86-min Kotori v2 set) next to ZS (zero-shot with the same reference) · identical short & long utterances · reference = real Kotori game voice
GSV-v4 GPT-SoVITS v4 (LoRA)GSV-v2Pro GPT-SoVITS v2Pro (SV)GSV-v2ProPlus GPT-SoVITS v2ProPlus (wide SV)Fish Fish-Speech S2-ProStyle-BERT Style-BERT-VITS2 JP-ExtraFT = fine-tuned on Kotori dataZS = zero-shot with same reference clip