Find a better source moment before enlarging it
If a paused video looks blurred, nearby moments may contain a clearer expression or a steadier camera view. Select a short range around the moment you need and capture several candidates. This page starts with a quarter-second custom interval to help compare a brief action. Review the subject at the same size across candidates rather than assuming the largest download or highest number is the best image.
The score is an estimate, not an AI verdict
This release uses a Laplacian edge-variation estimate. It is not an AI model for selecting the best frame, and it does not reconstruct missing detail. Texture, noise and sharp background text can increase the score even when the main subject is blurred. The gallery marks top estimates, including ties, as a convenience. You still choose the most useful picture.
Treat blur according to its cause
Motion blur can require a different source moment. Missed focus may affect the entire useful range. A low-resolution recording lacks fine detail even when edges look crisp, and compression can create false edges. Capture a cleaner candidate first, save PNG to avoid another lossy encode, then try the separate image upscaler if a larger output is useful. AI enlargement estimates detail and cannot establish that an inferred face, letter or texture was present in the original footage.
Questions for this workflow
Does this use AI to repair blurry frames?
No. Frame capture and its sharpness score are separate from the AI image upscaler. Neither promises recovery of missing original detail.
Why is the highest-scoring frame not my favorite?
The score measures edge variation, not expression, composition or subject relevance. Inspect the actual subject before choosing.
Continue your workflow
Read practical guides · Enlarge an image · Report a reproducible problem
Capture uses the shared frame-extractor limits: 2 GiB source, 8,294,400 decoded pixels, 120 pictures and 96 MiB retained images. Start with a short playable sample; codec support and device memory vary.