AI Short Drama Shot Generator

What Problem Does It Solve

In AI short drama production, transforming an abstract script beat (e.g., emotional conflict, character confrontation) into a specific vertical shot adhering to cinematic language is a common technical pain point. Manual storyboarding is inefficient, and generic video generation tools often lack precise control over shot composition, blocking, and performance rhythm, making it hard to directly produce usable, cinematic-quality final shots. This skill aims to automate the conversion from "text beat" to "deliverable shot."

Core Workflow

The skill's core function is to take a finalized dramatic beat and output a cinematic vertical micro-drama shot. Its workflow revolves around these key steps:

1. Construct Shot Card: It first parses the input dramatic beat text (containing character, action, scene, emotion, shot, etc.) and can optionally incorporate actor or scene reference images. The skill builds a detailed "shot card," using bold to specify character features, costume, scene, emotion, cinematic language (e.g., close-up, wide shot), blocking, composition, and visible outcome.
2. Select Generation Route: Based on the input type, the skill automatically or with confirmation selects the Beatra API call path:
   • Pure text beat → beatra.videos.generate
   • One strictly confirmed opening image → beatra.videos.animate
   • Strictly confirmed opening and ending images → beatra.videos.interpolate
   • Loose actor/scene references → beatra.videos.generate_from_references
3. Execution and Task Management: It submits the generation request via the bundled scripts/mcp_client.py. Before submission, the shot card content is frozen for user confirmation. After submission, the skill continuously polls the task status using beatra.tasks.get until completion and returns the video assets. Each submission constitutes an independent paid stage.
4. Review and Continuity: After generation, the skill reviews the returned video, checking actor features, cinematic language, emotional beat adherence, and dialogue rhythm against expectations, and handles subsequent revision requests (subject to payment and recovery rules).

Boundaries and Considerations

• Scope: This skill is exclusively for generating cinematic, live-action, or realistic human drama shots in vertical format for micro-dramas. It is not suited for illustration/comic-style dynamic manga, storyboard planning, or broad non-narrative videos; those requirements should be routed to other workflows.
• Payment and State Recovery: Each shot generation is a paid operation. Changing core elements like the beat, reference images, model, or aspect ratio constitutes a new paid job. If interrupted, recovery should use the identical stage ID and frozen payload; duplicate paid submissions must be strictly avoided.
• Model and Control: The default is model: "auto", but prior to generation, beatra.models.list should be used to confirm the model's capability matches the input media facts and shot control items.
• Client Constraints: All API calls must be executed through the bundled MCP client; one should not configure or use the host agent's Beatra Connector, nor resort to REST/OpenAPI fallbacks.