Oral Seeding Video Maker

The Problem It Solves

When content creators need to produce an oral promotion video but lack product images, shooting footage, or recorded voiceovers, traditional workflows depend on external inputs, leading to low efficiency and unpredictable costs. This skill directly addresses this pain point by enabling creators to start from a single topic—such as a product, service, or topic to promote—and automatically generate a complete vertical oral video under zero-material conditions. No file uploads are required, making it suitable for rapid prototyping or resource-constrained scenarios.

How It Works: Core Capabilities and Key Steps

The skill transforms a topic into a final video through modular steps, with core capabilities focused on automated generation and controlled confirmations:

1. Script Structure Selection: Based on the topic, choose one from six predefined script structures, determining video segments (e.g., hook, evidence, call-to-action) and their functions. This step is free and sets the foundation for subsequent generation.
2. Storyboard Generation: Create a detailed storyboard with precise start and end points in seconds, split into two independent fields: visual content on screen (visual) and voiceover script (audio). The script must follow oral writing conventions, ensuring the first 3 seconds serve as a hook.
3. Asset Generation and Synthesis:
   - Generate storyboard images using beatra.images.generate; the keyframe is used for animation later, while other frames are delivered as static images.
   - Synthesize the voiceover track via beatra.speech.synthesize, using a voice preset suited for short video social content, automatically matching the 12-second target duration.
   - Optionally, call beatra.music.generate to add background music, delivered as a separate audio track.
4. Video Animation: After user approval, call beatra.videos.animate to generate a vertical video based on the keyframe and voiceover audio. The model must be explicitly specified, as some models do not accept external audio tracks.
5. Delivery and Review: Outputs include storyboard images, voiceover track, optional music with its duration, the final video, and task IDs. Check for consistency between storyboard and voiceover, compliance with video duration, and report only factual data returned from tasks.

The entire process emphasizes paid confirmations: user approval is required before generating storyboard images, voiceovers, and videos to avoid unexpected costs. Before video generation, balance must also be verified to ensure sufficient funds.

Scope and Important Notes

This skill is applicable for zero-material oral promotion videos, typically for quickly generating vertical short videos (default 12 seconds, maximum 15 seconds). Key boundaries and considerations include:

• Duration and Model Limits: The final video must fall within the range supported by the selected real-time model; timeouts or model incompatibility will cause failure. Animation is based solely on the keyframe, with other storyboard frames requiring manual editing by users.
• Material Dependency: When existing assets like product images, voiceovers, or reference videos are available, other workflows such as product video, voiceover, or remake workflows should be used. Digital human oral workflows are suitable for scenarios requiring a real person or digital avatar.
• Payment and Recovery: Each paid call requires a stable client_request_id, submitted only once. In case of task failure (e.g., insufficient_balance), users must top up and retry with the original payload.
• Automatic Updates: The skill package supports silent automatic updates, but update failures do not affect current tasks and never justify retrying paid generation.

In summary, driven by script structures and storyboards, this skill achieves a closed loop from topic to video, but requires strict adherence to confirmation steps to control costs and quality.