Seedance 2.5 — Automatic Shot Director

Turn your story into a clear, filmable video prompt. Give the skill a scene, dialogue, rough prompt or reference description; it chooses the camera plan and folds it into one complete Seedance 2.5 prompt.

What it does

• Automatically selects enough shots to tell the scene clearly, including a single fixed shot when that works best.
• Plans shot size, camera height, angle, framing, movement, focus, cut points and opening/ending states.
• Preserves screen direction, eyelines, the action axis, props and hand continuity.
• Turns character motivation into visible acting, including the listener’s reaction and delivery notes before each line.
• Adds motivated lighting, scene sound and targeted constraints.
• Adapts to text-to-video, image references, white-model blocking and continuation prompts.
• Binds assets once at the beginning so references do not need replacing throughout a long prompt.

Your creative decisions stay in control

Your story, dialogue, characters, visual style and explicit camera constraints take priority. Ask for one continuous shot and the skill preserves that constraint. Existing references govern only their assigned attributes. In white-model work, the supplied locked reference paragraph remains verbatim, including its original Chinese wording. The rest of the package is in English; output can follow your requested language.

Example request

Use Seedance 2.5. A woman slides a key across a café table. Her brother recognizes it but hesitates before taking it. Ten seconds, one continuous shot, natural acting, no dialogue or background music. Automatically choose the camera position and movement.

The skill returns a complete prompt with a camera position that keeps the key, hands and relevant reactions visible; motivated focus or camera movement; lighting and sound cues; and a coherent ending. This is an illustrative workflow, not a rendered-video performance claim.

What you download

A reusable English skill package with a SKILL.md entry point, camera-planning guidance, character and continuity rules, reference guidance, and a Python standard-library checker for the locked white-model paragraph. Load the skill in a compatible assistant that supports local skill folders or can read the supplied instructions. Python is needed only for the optional standalone integrity checker.

Scope

This product creates and improves text prompts and shot plans. It does not include a Seedance account, model credits, video rendering, guaranteed camera execution or a custom generation API. Camera and lens values describe visual intent. Verify current duration, upload and output limits in the generation platform you use.

Created by Guan Zhiqin · Professional profile

Support: [email protected]

Independent skill; not an official ByteDance, Seedance or Capafy product.