Sony patents AI that predicts the next game frame before the GPU finishes rendering

What happened?

Sony Interactive Entertainment filed a patent for an artificial intelligence system capable of predicting and displaying the next frame of a game before the GPU finishes rendering it. The document, published on October 15, 2026 as US 2026/0320740 A1 and titled “Generation Super Sampling,” describes an encoder-decoder model that uses the already-rendered frame, motion vectors and, optionally, recent controller inputs to generate a predicted image.

Unlike current frame generation techniques such as DLSS or FSR, which typically wait for two complete frames to interpolate an intermediate one and add latency, Sony’s approach extrapolates from a single available frame. While the GPU renders the actual next frame, the AI processes the previous one and sends the prediction to the display buffer. The system can also predict multiple future frames and includes a non-AI fallback based on shifting pixel blocks.

Why it matters

The main drawback of current frame generation is input latency. By incorporating controller data (buttons and analog sticks within a short time window), the patent aims to anticipate player actions — similar to the principle of rollback netcode in fighting games. This could reduce stuttering in heavy scenes without the delay typical of existing solutions. The technology aligns with Sony’s AI upscaling work (PSSR/QSSR) and may be relevant for PlayStation 6, although the patent names no specific hardware or timeline.

Training uses consecutive frames from the game’s own video as reference, without human labeling. There are no performance or visual quality numbers in the patent; it remains an unvalidated technical proposal.

What changes in practice?

If successfully implemented, the technique could enable more stable frame rates on consoles without high-power dedicated frame-generation hardware while preserving better responsiveness. Players would experience less hitching in complex scenes and less disconnect between input and image. However, prediction accuracy, visual artifacts and the real-time computational cost of the AI still need to be demonstrated. Sony continues developing graphics technologies with AMD (Project Amethyst), and this is another sign of investment in AI for future consoles.

By GeekikiBot