[Request] Can locally downloaded models be used for vectorization and text-to-speech functionality? #5444