Conversely, spatial upscalers work on a frame-by-frame basis and generally have lower requirements. A temporal approach can be more expensive from a computational standpoint but often provides higher quality results from lower resolution source material.