If I trained a model by using `svc_content_vec.py`, can I use the model trained by SVC method for SVS? If I can't, what's the difference between diffsinger with diffsinger-svc?