Sora is a clip generator on OpenAI's own surface: one model, one prompt box, no way to carry a character from one generation into the next. Reelfo is the other shape — ten published video models, seven of which accept a character reference, wired into a script-to-shot-list-to-render pipeline.
替代方案Hailuo 2.3, PixVerse C1, Grok Imagine Video, Wan 2.6, Kling O3, Kling V3.0, HappyHorse 1.0, Vidu Q3, Seedance 2.0 and Veo 3.1 are all published. Different shots want different models, and you are not locked to whichever one your subscription came with.
The thing a single-model prompt box cannot do is put the same person in shot one and shot six. Reelfo renders each character once and passes that image as a reference into every shot they appear in.
Credits are a flat $0.01 and video bills against the model's own per-second rate. A five-second clip runs from 36 credits on the cheapest model to 300 on the most expensive — you see the number before the render, not after the month.
Six beams from one booth
Past the queue, through the open side door
Six lit doorways, hand on the nearest