magic-hour-text-to-video
Magic Hour text to video
Translate the idea into one clear visual event. Accept “make an elevator open onto the moon”; choose the framing, pacing and model within the user's existing budget. Preserve requested creative choices. Deliver the video, with any unmet requirement made explicit.
Decide whether a reference is necessary
Use text_to_video_create_video for an original scene whose exact cast and appearance can vary. If the user provides their face, a recurring character or a required starting composition, use that reference through AI Image Editor when needed, then Image-to-Video. An optional imagined character is not a reason to force a paid image step. Honor an explicit Text-to-Video request while explaining any identity limit that matters.
Direct a visible event
Define the setup, the change and the payoff. For a short clip, choose one event that a viewer can understand without explanatory copy. Establish the subject immediately; reserve enough time to see the result. Give physical direction and timing instead of piling up adjectives.
Examples of structures, not measured viral formats:
| Intent | Shot direction |
|---|---|
| Impossible reveal | Familiar closed door → it opens onto an unexpected world → hold long enough to read the reveal |
| Visual joke | Establish an ordinary action → one clear physical contradiction → visible reaction or consequence |
| Cinematic atmosphere | One focal subject, one camera move and restrained environmental motion; no unnecessary plot |