SOUTH+BRIDGE
Tech AI-translated

Smoothness Is Not Responsiveness

Frames inserted by AI can fool the eyes, but not the fingers. The feel of play diverges between generated 60 fps and computed 60 fps.

1UP · June 6, 2026 · 5 min read

AI Summary

The article argues that AI-based frame generation—such as NVIDIA DLSS 3's Frame Generation—inflates frame counts while adding input latency, creating a fundamental disconnect between what players see and what they feel. While the technology works reasonably well in slow-paced single-player games, it is harmful in fast-reaction genres like fighting games where frame-precise inputs are decisive. The author closes with a note of encouragement for small indie teams in Busan: responsiveness, unlike visual polish, is built through design rather than expensive hardware.

Smoothness Is Not Responsiveness

Fingers Don't Read Graphs

Turn on frame generation and the numbers double. 30 becomes 60; 60 becomes 120. The counter in the corner of the monitor climbs honestly. But the hands gripping the controller don't trust those numbers. Fire up NVIDIA DLSS 3's Frame Generation for the first time and sprint through Cyberpunk 2077, and the screen does flow more smoothly—pan the camera and ghosting fades, the landscape glides. Yet when an enemy dashes right in front of you, a faint lag appears between the moment you press the dodge button and the moment your character actually moves. Your eyes see 60 while your hands are chained to 30. That is the true nature of generated smoothness.

The reason is obvious once you look under the hood. Interpolation-based frame generation works by having AI insert a middle frame between two already-rendered frames. To insert that middle frame, the system needs the subsequent frame to exist first, so it holds the completed frame and waits. That wait is the latency. The display grows richer while the time from input to response actually increases. The 60 shown on the counter is a visual fact; the 30 felt by your fingers is a responsiveness fact. They are different numbers.

Games Get Caught Because They're Not Movies

This is where movies and games diverge. TV motion interpolation—the so-called soap opera effect—repels cinephiles, but at least in a film there is no input. The audience only watches. There is no way to feel with your hands the 0.05-second delay created by generated intermediate frames, so the only issue is picture quality. Games are different. They are not a medium you watch but one you touch. A closed loop runs between the screen and your fingers, and the round-trip time of that loop is the very skeleton of fun.

This is why fighting games cling to 60 fps like a creed. One frame in Street Fighter 6 is roughly 16.6 milliseconds, and the outcome of a match turns on frame-count differences in move startup. A 3-frame parry and a 4-frame parry are entirely different techniques. Even if you insert interpolated frames to make it look like 120 fps, input judgment still runs on the original computed frames—all you add is latency. A fighting game that looks smoother but responds more slowly is strictly worse. Nintendo's insistence on preserving the internal computational frame rate in competitive titles like Splatoon and Smash Bros. is not superstition; it is design ethics. Responsiveness is a non-negotiable floor; smoothness is a luxury placed on top.

The counterargument deserves equal attention. Not every game is a fighting game. In slow-camera, generous-input-window single-player adventures like Horizon or God of War, generated frames function almost like a free lunch: scenery looks better and latency stays below the perceptual threshold. For players who lack the graphics card to push native 4K at 60 fps, this is a genuine gift. Dismissing frame generation wholesale as a scam is lazy. The problem is not the technology but its application without regard for genre. The knife is not bad; the problem is trying to eat soup with it.

Smoothness as Marketing, Friction as Design

Upscaling shares the same trap. DLSS and FSR render at a lower resolution and use AI to fill in pixels to simulate 4K. Still images are nearly flawless. But on fast-moving thin objects—cables, hair, lattice patterns—flickering and ghosting seep through. This is a shared weakness of generative technology. Because the task is to plausibly fabricate information that was never there, errors are largest precisely when information density is highest—that is, during the fastest and most critical moments. In games, the fastest moments are usually the most decisive: the instant a boss changes its attack pattern, the instant an opponent closes in. Those are exactly the moments when generation falters.

I think this trend exposes a disease in how games are perceived. Frame counts, resolution, content volume, buzz-worthiness—the habit of ranking games by measurable quantities. Frame generation is a technology perfectly calibrated to that habit, because it stretches the benchmark bar graph further. But fun does not live in bar graphs. Fun comes from the trust you feel when the button you pressed returns a response at exactly that moment. Elden Ring is brutally beloved not because it has many frames, but because its dodge invincibility frames are honest. Friction, constraint, and precise response—that is where the feel of play lives.

For those crafting games with small teams in Busan, this is almost a comfort. Even without the budget to buy generated smoothness through the latest graphics card, responsiveness is made through design, not money. Cutting input latency by a single frame, coding hit detection honestly, tuning the dodge window to fit a player's fingers—none of that requires expensive AI. A game with fewer pixels can still feel glued to your hands. Before envying generated 60 fps, the priority is making computed 30 fps accurate. Smoothness is a luxury good; responsiveness is a necessity. What indie developers must sell has always been the latter.

This article was automatically translated from the Korean original by AI. For the authoritative version, read it in Korean.

한국어 원문 읽기 →