15s Vertical Short Video (text-to-video)

When to use

The user asks for a short vertical clip generated from an idea, with no source footage and no reference image. Deliverable = one playable video URL.

Method (order matters)

1. Intent first. Decide the single piece of information / emotion this shot carries, then the shot size (extreme wide → close-up = information dial). Never start from visual adjectives.
2. One visual centre. Exactly one dominant element per frame; everything else supports it. If squinting at a thumbnail gives two focal points, rewrite.
3. Lighting with a motive. Write the key:fill ratio explicitly and give the light an in-frame source (window, practical, neon).
4. One movement per shot. At most one camera move; lock it with "static" if the shot should hold.
5. Prompt skeleton (camera first): camera movement → subject + action → environment → lighting/atmosphere → style/technical.
6. Cheapest acceptable render: provider minimax-h3, duration=15, resolution=768P (about $1.04 per clip). Do not silently upgrade the tier; a higher tier is a separate, user-confirmed decision.

Call contract

Call beehive_submit_video exactly once with:

field	value
prompt	the five-part skeleton assembled above, in English
duration	15
resolution	768P
ratio	9:16

Hand-off

Report the artifact URL verbatim (dest_video_url) plus the job id. If the job fails, report the platform error verbatim — never substitute a different clip or a "closest" result.