Intel Video Processing Library Adding AI Assisted Video Encoder Features

Phoronix
Intel's libvpl 2.16 introduces experimental APIs for AI-assisted video encoding, including saliency and ML-based adaptive target usage controls.

Summary

Intel has released version 2.16 of its Video Processing Library (libvpl), which primarily introduces experimental Application Programming Interfaces (APIs) for AI-assisted video encoding functionality. Libvpl acts as a dispatcher between the Intel Media SDK and the VPL GPU Runtime components. The key addition is the new `mfxExtAIEncCtrl` interface, which is experimental and designed to control AI-assisted encoder solutions. Specifically, the new APIs support a saliency encoder solution and an ML-based adaptive target usage solution. The author notes that while the header structures for these features are present in the open-source VPL GPU Runtime code, there are no immediate associated commits showing the actual implementation, suggesting these features might be reserved for binary-only components. The release is available on GitHub.

(Source:Phoronix)