Freeze the authorized script and its non-negotiable meaning

Record the script owner, version, permission scope, confidential handling, intended audience, destination, duration, and approving editor. Mark dialogue, chronology, character knowledge, factual claims, required actions, locations, props, emotional turns, and material that must not be visualized. Separate literal source obligations from interpretive choices. Resolve ambiguity with the writer or label it for review; do not silently invent connective events. For Seedance 2.5 Script to Video, the first deliverable is a traceable beat table, not a generated clip. Each row should point back to the exact source span and explain what the viewer must understand after the beat.

• Script version and authorization
• Beat-to-source pointer
• Meaning and factual obligations
• Permitted adaptation choices

Turn approved beats into economical shot contracts

Assign every selected beat a narrative job, subject, action, setting, framing, camera behavior, duration range, audio role, continuity dependencies, reference assets, and rejection conditions. Combine beats only when the dramatic turn remains legible; split them when one shot would hide two consequential actions. Track character appearance, wardrobe, objects, screen direction, location state, dialogue timing, and evidence obligations across the sequence. Use visual references only within their licensed scope. A shot contract should tell an editor why the material belongs in the cut and what may change during revision, instead of relying on a broad cinematic prompt to preserve story logic.

Verify the named product context and review the complete cut

Before submitting material, record whether the current authorized surface actually shows the Seedance 2.5 label, accepted script or prompt inputs, reference controls, duration options, audio handling, region, entitlement, terms, and export behavior. If the path is absent, mark the attempt blocked and keep the production packet portable. Preserve exact inputs, settings, outputs, failures, and revision reasons. Assemble candidates with temporary audio and captions, then compare the cut with the source for changed motivation, omissions, invented facts, continuity breaks, identity drift, dialogue errors, pacing, and delivery fit. The result is a bounded adaptation decision, not proof of model capability or script-to-video automation.