The demand for AI-generated video has exploded in 2026, but most teams hit the same wall: they can generate a clip, but keeping a character or product looking consistent across scenes is nearly impossible. Seedance 2.0 Fast solves this with something I haven't seen in any other video model at this price point: a direct reference image system where you tag your inputs by name and control exactly which reference appears where in the video. That capability alone changes what's possible for...