Interpolation
Traditional resizing estimates new pixels from existing neighboring pixels.
VIDEO SUPER-RESOLUTION
Video super-resolution.
AI upscaling beyond simple image resizing.
Video super-resolution generates higher-resolution frames from lower-resolution input. For video, quality also has a time dimension: details should remain stable as objects move from frame to frame.
How it differs from ordinary resizing
AI super-resolution
A learned model generates higher-resolution frame content. Remaster AI runs trusted GPU processing profiles rather than arbitrary browser-selected runtime assets.
Temporal behavior
A frame can look sharp in isolation but still flicker or produce unstable detail during playback.
Current Remaster AI profiles
Eligible input at 756 px wide or less is classified as SD→FHD. Eligible input at 1368 px wide or more is classified as FHD→4K. The 757–1367 px range is outside the current standard profiles.
Important limitation
Super-resolution does not prove what was present in an unclear source. Generated visual detail should not be treated as guaranteed factual recovery of text, faces or other missing information.