Upscaling
Upscaling is computing an image or film to a higher resolution. Modern methods add plausible detail rather than merely multiplying pixels. They get visibly more out of what exists, but they do not invent what was never there.
Why it belongs to AI production
Generative video models often deliver lower resolutions than the final size requires, because computing time rises with area. A film to be delivered in 4K is therefore produced smaller and computed up at the end. That is not a workaround but the usual route, as long as it happens in the last step and not in between.
How to spot bad upscaling
- Waxy skin: pores and fine structures get smoothed, faces look artificial.
- Flickering edges: the additions differ from frame to frame, which shows as unrest.
- Invented detail: type that was illegible suddenly spells letters that were never there.
So we check upscaled material at full resolution and in motion, not only as a still. Type is laid over in the edit after upscaling anyway.
Frequently asked questions
What is upscaling?
Upscaling describes computing image or video to a higher resolution, for instance from Full HD to 4K. Modern methods add plausible detail, but they cannot restore information that is missing.
Do you deliver in 4K?
Yes, where it is needed, for trade fairs and large screens for example. For website and social media Full HD is usually sufficient and saves loading time. What is delivered is stated in the offer.
Does upscaling make a bad image good?
No. It improves resolution, not the idea, the light or the sharpness of the capture. Where the source material is unusable we produce the shot again.
Related terms
Read on: All terms in the glossary • Pricing
Trade fair, cinema or website?
Where the film runs decides resolution and bitrate. Tell us the destination.
Book a first call