Kling v3 Video Generation

Problem

Video generation workflows often get stuck on practical plumbing: API keys, task polling, local asset upload, and error recovery. A caller must prepare prompts, handle image and video fields, interpret async responses, and deal with balance or authentication failures. kling v3 wraps the hosted Kling v3 interface behind dLazy into a command-line client, allowing agents to start text-to-video or image-to-video jobs with consistent parameters and continue based on the returned state. It consolidates trigger keywords, authentication, output format, and error handling into one entry point, reducing the cost of stitching multiple API scripts together.

How it works and limits

The core path is: a user provides a dLazy API key, and the skill sends prompts, reference images, and generation parameters to api.dlazy.com through the dlazy CLI. Local files are uploaded to dLazy media storage so the model can read them, while generated results are returned as URLs on files.dlazy.com. With --no-wait, the response includes generateId and status, requiring follow-up polling. This skill fits automated pipelines that need reliable Kling v3 calls, async task handling, error-code mapping, and result links. It is not a local offline inference tool and does not bypass dLazy service terms. For teams that need call-chain auditability, input/output retention, or sensitive asset handling, the cloud storage and API data boundary should be evaluated carefully. Errors such as insufficient balance or missing credentials should be surfaced clearly to the user, with a next step to fix the credential or credit issue, rather than silently retried.