Diffusion steps

A diffusion model produces an output by starting from random noise and iteratively denoising over N 'steps'. More steps typically mean higher detail and quality, at the cost of more compute and generation time. Fast variants of models (like Veo 3.1 Fast or LTX 2) use fewer steps; flagship variants (Sora 2 Pro, Veo 3.1 full) use more. Most users don't need to think about step counts — they're tuned by the app.