TrueFrame

Cloud media platforms, post-production studios, and digital archives face a significant cost and performance bottleneck in delivering premium-quality video. Achieving "Netflix-grade" encoding quality typically requires the same video to be encoded multiple times using different settings before selecting the optimal result. This computationally intensive trial-and-error process increases processing time, cloud computing costs, energy consumption, and operational overheads across large-scale video workflows.

TrueFrame eliminates this inefficiency by using its AI-powered LiteVPNet engine to predict the optimal AV1 encoder settings in a single pass. First demonstrated at IEEE PCS 2025, the hardware-agnostic software development kit (SDK) reproduces optimal encoding quality in 87.3% of video clips on the first attempt, delivering a 65-fold speed improvement over conventional optimisation workflows. For the small proportion of clips requiring additional refinement, only a second encoding pass is typically needed, preserving substantial efficiency gains.

Designed as a sidecar solution for existing GPU-based encoding infrastructure, including NVIDIA platforms, LiteVPNet integrates seamlessly with current video production pipelines without requiring changes to downstream playback systems. This project will advance the technology to a partner-validated, licensable product (TRL 7) while establishing the commercial foundations for a Trinity College Dublin spin-out. Addressing a combined market opportunity exceeding €10 billion, TrueFrame enables media technology providers to reduce encoding costs, accelerate content delivery, and significantly lower the energy and carbon footprint of large-scale video processing.