Stability.ai and probably others too already working on video and audio models too and also I think I heard a service to train / finetune the model with your own dataset.