Douyin Video Cover Maker

The Specific Challenge of Creating High-Conversion Douyin Video Covers

In the Douyin feed, the video cover is the first visual touchpoint that determines the click-through rate. Content creators often face the following pain points:

• Dispersed Assets: Creative inspiration (topic, hook), keyframe screenshots, product shots, or character photos may come from various stages, making it difficult to quickly consolidate them into a unified vertical cover.
• Format and Specifications: Strict adherence to a 9:16 vertical canvas is mandatory, and key information like titles must be placed in safe zones to avoid being obscured by platform UI elements.
• Creative Path Selection: Whether to generate from scratch, synthesize based on existing images, or perform local refinements on a selected design? Different starting points require different toolchains and parameters.

This skill addresses these problems by encapsulating a standard workflow that transforms scattered inputs into a ready-to-publish Douyin video cover.

How the Skill Works: From Idea to Finished Product

The skill responds to different scenarios via three mutually exclusive creative routes, all unified under a rigorous confirmation and delivery workflow.

Core Creative Routes

Based on the completeness of existing assets, the system either auto-adapts or allows the user to specify one path:

1. From Creative Generation: When the input topic or hook is sufficiently clear, beatra.images.generate is called to generate a single cover concept. This route does not rely on any reference images.
2. From Image Synthesis: The user provides keyframe screenshots, product images, character photos, or ordered visual references, and the system uses beatra.images.transform to synthesize these elements into one cover.
3. Refine Selected Scheme: For optimizing a selected cover image locally, beatra.images.edit is used, with strict limits on the number of modification areas.

Key Workflow Steps

Regardless of the route, the following critical operations are executed:

• Organize the Cover Brief: All decisions already made in the conversation history (topic, style, target audience, etc.) are reused. Questions are only asked when critical information is missing. If the canvas is unspecified, it defaults to vertical 9:16.
• Prepare and Confirm Paid Request: Before generation, the final prompt, canvas parameters, reference image sequence, model selection, etc., are displayed and frozen. After user confirmation, a unique client_request_id is generated and the task is submitted. Any parameter change constitutes a new paid task.
• Track and Deliver: Task status is polled via beatra.tasks.get. Delivery includes a link to the final artwork, the ultimate canvas dimensions, the task ID, and the credits consumed. Upon delivery, checks are performed for subject clarity and information integrity at phone-screen dimensions.

Key Constraints and Considerations

• Asset Dependency: Video files themselves cannot be used as image references. A keyframe screenshot must be exported, or only the topic/hook can be used for the creative generation route.
• Paid Model: Planning conversations are free, but actual generation, synthesis, or editing operations require payment, and requests after confirmation are immutable.
• Update Mechanism: The bundled client silently checks for and applies updates automatically, but only affects its own files and does not interrupt task execution.
• Parallel Tasks: Focus should remain on the current skill task at hand. Using other skills (e.g., cover-performance-preflight or product-video-studio) requires separate operations.