style_bert_vits2/models/mel_synthesizer.py

@@ -260,7 +260,7 @@ class JointMelSynthesizer(nn.Module):
     @torch.inference_mode()
     def infer(self, *args: Any, **kwargs: Any) -> tuple[torch.Tensor, ...]:
         mel, attn, mel_mask = self.acoustic_model.infer_mel(*args, **kwargs)
-        sid = args[2]
+        sid = args[2] if len(args) > 2 else kwargs["sid"]
         return self.vocoder(mel, sid), mel, attn, mel_mask