# Legacy timed and bracketed scene modes

## Legacy timed 2.0 style
Use a compact world/identity block followed by contiguous timed action beats, camera, light, sound and an avoid line. Collect duration, music/no-music and English/translated output preference together when missing. For an optionally translated block, all prose should be in the requested language while reference identifiers, timestamps and technical placeholders remain unchanged. Put the English explanation outside the copy-ready translated block.

## Bracketed scene 2.5 style
Use [Global Setting], [Character Styling], [Core Performance], [Timestamp Storyboard] and [Negative Prompts]. The labels are a useful house format, not a guaranteed API schema. Reference roles precede story action. Each material has a function and an exclusion: appearance only; do not inherit the room, clothing or extra people. Map actual attachments in order, never guessed local names. Use a short staged continuity block only for a complex multi-character or multi-location scene.

## Mode-specific extensions
For an edit, identify the original clip, exact change and retained properties. For forward/backward extension, identify the boundary state and the direction of time; keep motion and lighting continuous. For first/last frames, compare endpoints and describe a feasible transition rather than reinterpreting both. For a storyboard grid or blockout, distinguish composition guidance from final appearance. For many references, select only the needed materials per scene and keep their roles unambiguous. The current tool decides what is supported.

## Action and camera discipline
Choreograph observable motion with path, contact and ending state. Separate one camera move from subject action. Rotate framing only when a cut helps meaning; a requested one-take/locked portrait keeps its framing. A final hold provides an editing boundary, but do not automatically reserve three seconds if the brief is shorter. Re-establish every standalone clip’s subjects, outfit, place and object states. Do not depend on hidden memory, private applications, outside skills or a deck parser.

## Sound and translation
Separate dialogue, music, effects and on-screen text so unsupported syntax can be removed. Preserve the exact approved spoken words and language. Do not require invented bracket/control syntax as a universal platform feature. Default no music only if the user requested it; otherwise clarify or state the assumption. Translate requested prose faithfully and verify timestamps/reference IDs remain identical.