Low-complexity On-device Super Resolution

A machine-learning based, low-complexity super resolution algorithm for real-time high-definition video. It consistently delivers image quality on par with open-source deep neural network methods, while industry-leading optimization dramatically increases execution speed: it covers more than 80% of mobile devices at low power, staying cool and quiet. Our AI inference engine, purpose-built for video processing, delivers more than 20 times the performance of native inference engines.

Key Features

  • Kirin 980 / Snapdragon 845: 1080p → 4K at 40 fps; Core i9: 4K → 8K at 1000 fps
  • Works across all content types with no failure cases
  • A truly practical on-device super resolution solution

Tell us about your project

Contact us