SINE: SINgle Image Editing with Text-to-Image Diffusion Models