Scene Schema

Scene Schema

Scenes appear in two main places:

• AnalysisResult.scenes[] — Output from video analysis.
• ScenarioScene[] — Output from scenario generation and input to image/video generation.

UserIntent (scenario input)

Defined in packages/shared/src/types/scenario.ts as UserIntentSchema:

export const UserIntentSchema = z.object({
  mode: z.enum(["style_transfer", "content_remix"]),
  product_name: z.string().optional(),
  product_description: z.string().optional(),
  target_audience: z.string().optional(),
  language: z.string().optional(),
  video_duration: z.number().optional(),
  scene_count: z.number().optional(),
});

Field	Type	Required	Description
mode	"style_transfer" | "content_remix"	Yes	Video generation mode.
product_name	string	No	Product or subject name.
product_description	string	No	Description of the product/subject.
target_audience	string	No	Target audience.
language	string	No	Output language code.
video_duration	number	No	Total video duration in seconds.
scene_count	number	No	Number of scenes to generate.

ScenarioScene (scenario output)

Defined in ScenarioSceneSchema in packages/shared/src/types/scenario.ts and mirrored in the OpenAPI ScenarioResponse:

export const ScenarioSceneSchema = z.object({
  scene_index: z.number(),
  duration_seconds: z.number(),
  image_prompt: z.string(),
  negative_prompt: z.string().nullable().default(""),
  video_prompt: z.string(),
  text_overlay: z.string().nullable().optional(),
  transition: z.string().nullable().default("cut"),
});

Field	Type	Required	Description
scene_index	number	Yes	0-based scene index.
duration_seconds	number	Yes	Scene duration in seconds.
image_prompt	string	Yes	Prompt for image generation.
negative_prompt	string	No	What to avoid in the image (may be "").
video_prompt	string	Yes	Motion/camera prompt for image-to-video.
text_overlay	string	No	Optional on-screen text for the scene.
transition	string	No	Transition type to the next scene (e.g. "cut", "fade", "dissolve").

AnalysisResult scenes (analysis output)

In ARCHITECTURE.md and the OpenAPI ScenarioAnalysisInput, analysis scenes look like:

Field	Type	Description
index	number	1-based scene index.
duration_seconds	number	Scene duration.
description	string	Visual description of the scene.
transition	string	Transition type between scenes.
text_overlay	string	Text shown in the scene, if any.
camera_movement	string	Camera movement (e.g. "pan left", "zoom in").

These analysis scenes can be passed into scenario generation (as analysis) to influence the resulting ScenarioScene array.

Relationship in the pipeline

• Analyze step → returns AnalysisResult.scenes[] with structural info.
• Scenario step → takes UserIntent and optional analysis.scenes[] to produce ScenarioScene[].
• Generate step → uses ScenarioScene.image_prompt, negative_prompt, video_prompt, and duration_seconds for image and video generation.

See also

• Pipeline Overview
• UI Guide: Scenario
• AI Endpoints