ASAysha Shafiq
VIDEO SYSTEMS · GENERATIVE AI

VIDER Generative Video Reconstruction

I led a model-agnostic streaming framework that sends sparse keyframes and reconstructs the missing video with GPU compute near the viewer.

97.69% avg. bandwidth savings
0.83 avg. ViSiL similarity
28-video evaluation
Beat H.264 CRF 50 on savings + similarity

Trading bandwidth for compute

VIDER sends sparse keyframes and spends GPU compute near the viewer to regenerate the rest.

Model-agnostic pipeline

I led a keyframe-plus-interpolation design whose transport does not depend on any one generative model.

Evaluation

Evaluated across six content categories against traditional codec behavior at matched bandwidth.

Food Crisis Forecasting with Urdu NewsBookSwitch Browser ExtensionSlack AI Course Bot