Sony has filed a patent for an AI-driven technology designed to predict upcoming game frames before the GPU completes rendering. The system aims to reduce visual stuttering and minimize latency issues typically associated with frame generation techniques.
The technology functions by bypassing the need to wait for two completed frames to generate an intermediate one. Instead, it uses a jointly trained codec to convert rendered images into compact numerical representations, allowing the decoder to predict the next frame independently.
The AI system operates in parallel with the graphics pipeline. While the hardware renders the next frame, the AI processes the previous one and sends the prediction to the display buffer, potentially allowing the predicted image to appear before the actual render is finished.
To improve accuracy, the system incorporates motion vectors and recent controller input data. The neural network is trained on continuous video sequences, comparing its predictions against actual frames to calculate and minimize error rates.
