Doubao Image and Video Generation

Context and Problem

When building applications on top of the Doubao (豆包) model, developers often need to integrate its powerful capabilities for text-to-image generation, image transformation, and video synthesis. Calling the underlying APIs directly necessitates handling complex request schemas, managing asynchronous video generation tasks, and diagnosing errors like InvalidEndpointOrModel.NotFound. This requires deep familiarity with the Volcengine Ark endpoint model and additional code to encapsulate these common operations.

Core Capabilities and Key Workflow

This native skill streamlines interaction with the Doubao media generation API through pre-wrapped commands.

1. Media Content Generation
The skill provides distinct sub-commands for different use cases:
* Image Generation: Supports both pure text-to-image creation and image-to-image generation that uses one or more reference images as input.
* Video Generation: Supports both text-to-video and image-to-video modes. It intelligently appends key parameters like --dur (duration), --fps (frame rate), --rs (resolution), and --ratio (aspect ratio) to your prompt if they are not already present.

2. Critical Input Rules
To ensure precise control, the skill follows these rules by default:
* When you have a provisioned Volcengine Ark endpoint, always prefer the --endpoint-id flag. It only falls back to using the model name when endpoint IDs are unavailable.
* If image URLs are supplied in the input, the skill passes them through exactly as-is without attempting to download or re-host them, guaranteeing a deterministic workflow.

3. Asynchronous Task Management and Troubleshooting
The video generation process is asynchronous. The skill’s design thoroughly accommodates this pattern:
* Upon successful initiation, it returns a task_id. You can use the query sub-command with this ID to check the task status and retrieve the result.
* The wait sub-command enables automatic polling until the task completes, with an option to download the resulting file, greatly simplifying the logic for waiting on async operations.
* For endpoint or model errors, the built-in troubleshooting logic prompts you to first verify your Volcengine Ark endpoint authorization, which is the most common root cause.

Applicable Boundaries and Considerations

It is crucial to understand the skill’s positioning and limits:
* It is a native skill optimized specifically for the Doubao model, not a general-purpose MCP service. You do not need to spin up an upstream MCP server unless you explicitly require MCP compatibility.
* Its full functionality is strictly dependent on the Volcengine Ark platform and valid endpoint authorization. Ensure your calling environment is correctly configured with the necessary credentials or endpoints.
* The video generation features mirror the upstream behavior of the Doubao model. For advanced customization beyond the skill’s preset parameters, you may need to consult the API documentation (references/api-notes.md) directly.
* The skill focuses solely on media generation and task querying; it does not cover content post-processing (e.g., advanced editing) or the scheduling of non-Doubao models.