Labsco
MCP SERVER

Ghibli Video

by MichaelYangjson

Turn a still image into an animated video, follow the job by its task id, and read the points left on the key you are spending.

Video Generation & EditingVerified
Summary
The call surface maps straight onto the workflow — image_to_video submits, get_task_result collects, get_points says what is left — and each one wants api_key handed in explicitly.

Passing the key as a tool parameter rather than reading it from the environment is the detail to weigh: the credential travels inside the call, so whatever holds the conversation holds the key. The endpoint is configurable through Ghibli_API_URL, which means the same server can be aimed at a host other than the one you assume. Steering is confined to prompt, negative_prompt and aspect_ratio — there is no seed, no duration and no model selector in the arguments, so reruns are shaped by the prompt alone.

What it is

A thin client for a Ghibli-style image-to-video service: submit an image with a prompt, poll the task, and check the remaining points.

What you get
  • image_to_video starts a conversion from an image, steered by prompt, negative_prompt and aspect_ratio.
  • get_task_result returns the outcome for a taskId, so submitting and collecting are separate calls.
  • get_points returns the points remaining for the key in use.
Requirements

The service endpoint in Ghibli_API_URL, and an API key — every call takes it as an api_key parameter rather than reading it from the environment.

Setup effort

One command — npx -y @openmcprouter/mcp-server-ghibli-video