IT lexicon › Programming › Frame generation

Frame generation

Programming På svenska → Updated: 2026-10-26

Weaving entirely synthetic frames in between the rendered ones — more frames per second with no extra rendering work, at the cost of added latency.

The technique broke through in 2022 with Nvidia's DLSS 3 on the RTX 40 cards, followed by AMD's FSR 3 and driver-level options such as AFMF. Using motion vectors and optical flow, an intermediate frame is computed between two already-rendered frames and slotted into the stream, so the frame counter roughly doubles.

The catch is that the generated frame sits between two real ones, so the engine must hold the later one back — latency rises rather than falls, and fast motion or UI elements can produce artefacts. It also needs a decent base framerate to feel good. Do not confuse it with the upscaling in DLSS and FSR, which reconstructs resolution; it is often paired with latency reducers like Nvidia Reflex.

← Back to the lexicon