Sony may be building its own take on AI frame generation, and a new patent shows it wants to fix the biggest complaint players have about the technology. Sony Interactive Entertainment had a patent application titled "Generation Super Sampling" published earlier this week (October 15th) in the United States. It describes a system that predicts upcoming frames from the image on screen, motion data, and even the buttons you just pressed. Sony has not announced the PlayStation 6, and a patent is not a product announcement, but it lines up closely with what Sony has already said about frame generation coming to PlayStation.
How Sony's Frame Prediction Works
The application, published as US 2026/0320740 A1, lists Sony engineer Denis Dupeyron as the inventor. It was filed this summer (June 23rd) as a continuation of an application Sony first filed in December 2023, which was published in June 2025.
The system is built around a pair of neural networks. An encoder looks at the frame that was just rendered and turns it into an embedding, a compact set of data describing what is in the image. A decoder then uses that embedding to predict what the next frame should look like. The prediction can also take in motion vectors, which show which way parts of the image are moving, along with recent player inputs such as button presses or joystick and mouse movement. The predicted frame goes to the display while the console's normal graphics pipeline is still working on the next real frame.
The filing also covers predicting several future frames at once, and it includes a simpler non-AI fallback that shifts blocks of pixels along their direction of motion. Sony frames the whole idea as a way to smooth out "hitching," the sudden stutters that show up when a scene gets too heavy, and to help a game hold a steady frame rate.
Why Your Inputs Matter
The patent goes straight at the latency problem. Current frame generation on PC, such as Nvidia's DLSS 3, slips a generated frame between two frames the GPU has already finished. The game looks smoother, but it has to wait for that second real frame before anything can be shown. Sony's application calls this out directly, noting that the approach "creates a lot of latency because the device must render the last frame and the prior frame" before the generated frame can be inserted.
Sony's version predicts forward instead of filling in a gap. If the system knows you just flicked the camera stick to the right or hit jump, it has a better idea of what the next image should contain before it is ever drawn. In theory, that would give players the smoother picture of frame generation without as much of the input delay that makes some players turn it off.
What Sony Has Actually Confirmed
Sony has already confirmed that machine learning frame generation is coming to PlayStation hardware. PS5 and PS5 Pro lead system architect Mark Cerny said back in March, in an interview with Digital Foundry, that frame generation is part of the company's work with AMD.
"FSR Frame Generation is also based on co-developed technology (or as my good friend Jack Huynh puts it, 'co-engineered technology'). I'm very happy with how that work is progressing, and an equivalent frame generation library should be seen at some point on PlayStation platforms," Cerny said. When asked whether it would land on the PS5 Pro or next-generation hardware, he only said, "All I can say is that we have no more releases planned for this year. And that I look forward to discussing this more in the future!"
What is not confirmed is that this specific patent will power that feature, or that it will be part of the PS6 at all. Cerny made that point himself when a different Sony upscaling patent made the rounds earlier this year. "We file many patents during the course of business, not all of which relate to released systems or future products," he said. Sony has also not shared a release window, specs, or price for its next console.
Project Amethyst Points The Same Way
The patent does fit the direction Sony and AMD laid out last year with Project Amethyst. In a video released in October 2025, Cerny and AMD's Jack Huynh previewed three technologies for future hardware: Neural Arrays, which let groups of GPU compute units work together on machine learning tasks, Radiance Cores, which are dedicated hardware for ray tracing and path tracing, and Universal Compression, which cuts down on memory bandwidth. Cerny said at the time that the technology was still early and aimed at a future console a few years out. Here is the full video:
The work has already reached current hardware. Earlier this year, the PS5 Pro got an upgraded version of PlayStation Spectral Super Resolution, and Cerny confirmed the new PSSR uses the same core algorithm co-developed for AMD's FSR Redstone upscaling. Some PS5 games already use AMD's older FSR 3 frame generation, but that version does not use machine learning.
The Next Xbox Is Heading The Same Way
Sony is not the only one betting on smarter rendering. At GDC in March, Microsoft confirmed that its next Xbox, Project Helix, runs on a custom AMD chip co-designed for the next generation of DirectX and FSR. Xbox's Jason Ronald said the console "integrates intelligence directly into the graphics and compute pipeline," and Microsoft plans to start shipping alpha hardware to developers in 2027. With RAM costs already pushing the PS6 price debate toward four figures, technology that gets more out of the same hardware could matter a lot next generation.
Would you turn on frame generation on the PS6 if it really cut down on input lag, or do you only want frames the console actually renders?
Let us know your thoughts in the comments below.