video-gen

What This Skill Does

The video-gen skill provides a robust interface for the Tencent Cloud VOD AIGC API, enabling OpenClaw agents to generate high-fidelity videos from text or static images. Whether you are creating cinematic B-roll, visual content for marketing, or experimenting with advanced generative models like Kling, Hailuo, or Vidu, this skill abstracts the complex API requests into simple command-line arguments. It handles model versioning, aspect ratios, resolutions, and advanced image-to-video transitions, making it an essential tool for automated content pipelines.

Installation

To integrate this skill into your environment, ensure you have the OpenClaw CLI installed. Execute the following command in your project terminal:

clawhub install openclaw/skills/skills/autogame-17/video-gen

After installation, configure your Tencent Cloud credentials in your environment file (.env) using the keys: VITE_VOD_SECRET_ID, VITE_VOD_SECRET_KEY, and VITE_VOD_SUB_APP_ID. Failure to set these will prevent the skill from authenticating with the cloud provider.

Use Cases

• Marketing & Ad Creative: Rapidly iterate through visual concepts by converting brand assets into dynamic 16:9 or 9:16 promotional videos.
• Content Automation: Integrate with Feishu or other messaging platforms to push automated render progress updates to your team using the --chat-id flag.
• Storyboard Animation: Use the Image-to-Video functionality combined with start and end frames to create smooth, professional morphing sequences for project mockups.

Example Prompts

1. "Generate a 1080P video of a bustling Tokyo street at midnight using Kling 2.5, and post the progress to my project chat."
2. "Take this image [https://site.com/hero.jpg] and animate the character walking across the screen using Hailuo 2.3 at 16:9 ratio."
3. "Create a smooth transformation from this desert sunset image to a snowy mountain range scene using Kling 2.1 start and end frames."

Tips & Limitations

• Resource Management: Ensure you check your cloud provider's quotas, as high-resolution (4K) generation with top-tier models like Kling o1 can consume significant API credits.
• Prompt Engineering: While the --enhance flag is available, providing descriptive, detailed prompts about lighting, motion, and subject style yields significantly better results across all supported models.
• Compatibility: Not every model supports all features; for example, ensure you are using the correct model version (like 2.1+) when utilizing the --last-frame parameter for video morphing, as older versions may ignore this input.