360 Secure SkillHub
360 安全技能中心

Douyin Video Cover Maker

该 Skill 通过捆绑的 MCP 客户端调用 Beatra 图像服务,把选题、钩子、脚本或参考图片转成一张竖版 9:16 抖音短视频封面:先冻结提示词、画布、参考图顺序、模型与数量并请求确认,再一次性提交付费生成,随后轮询任务、按手机尺寸复核并交付图片与计费信息。

分类:创作
beatra-ai
beatra-ai
热度 13.35
最近更新 2026-10-21 11:21:50

安装

发给 Agent
帮我安装这个技能:npx secure-skills-cli add beatra-ai/douyin-cover-maker
兼容性
通用兼容
联网要求
需要联网
数据出网
必须出网
Metadata
name
douyin-cover-maker
description
Create a Douyin cover or vertical short-video cover from a video topic, hook, script, portrait, product photo, or reference image. This AI cover maker builds a clear focal visual and headline-safe composition for Douyin short-video covers, creator series, product demos, tutorials, food, lifestyle, and knowledge content, then refines an accepted draft into a consistent account-cover direction.

Douyin Cover Maker

Create one Douyin or vertical short-video cover from a topic, hook, script, key-frame screenshot, product or portrait photo, visual reference, or accepted draft. It covers Douyin cover, Douyin short-video cover, viral-style cover, and vertical-video cover requests. Reuse decisions already present in the conversation and take the shortest route to one publish-ready cover.

Choose the route

  • Generate from an idea: turn a confirmed topic, hook, or script into a single cover concept with beatra.images.generate. Do not enter this route on a style word alone.
  • Compose from an image: upload a key-frame screenshot, product, portrait, or ordered visual references and use beatra.images.transform. A video file is not an image reference: when the user starts with a video, ask for one exported key frame or screenshot, or use the topic-only route.
  • Refine an accepted draft: use beatra.images.edit with the selected draft as images[0]; use at most two normalized local edit regions.

This Skill creates one new cover. Route an existing cover that only needs a performance review to cover-performance-preflight; route a full video build to product-video-studio.

Shape the cover brief

Reuse the stated topic, hook, account style, target viewer, visual references, and must-keep details. Ask only when a missing choice would materially change the result. Without a photo or key frame, the topic or hook is a hard input: do not invent one, and do not pay on a style word alone. For a Douyin feed cover with no stated canvas, default to vertical 9:16. Do not reuse a Xiaohongshu 3:4 default.

Build around one visual hook, one focal subject visible at phone size, one style direction, and any ordered image roles. When the user already wrote title copy, place it in the top or bottom third and keep it from filling the frame. Otherwise default to a text-safe area rather than promising rendered Chinese typography. If the user requests in-image words, include the exact short text and its third-of-frame placement in the confirmation and verify it only when the result is visible.

Prepare and confirm one paid request

Use only this Skill's bundled scripts/mcp_client.py for every remote Beatra operation. Do not configure or call a host Beatra Connector and do not use REST/OpenAPI as a fallback. Upload local image files through the bundled client, then pass their artifact references in the declared order. The upload does not inspect an image; review only facts the host can actually see.

Keep model: "auto" and count: 1 by default. Call beatra.models.list only for a real availability, compatibility, control, or price decision. Before the generation, show and freeze the final prompt, canvas, ordered references and roles, headline treatment, model, controls, and count. Planning is free; generation is not.

After approval, create one opaque stable client_request_id that contains no user content and submit exactly once. Any changed prompt, reference or order, canvas, model, count, or control is new paid work and needs a new confirmation and a new request ID. Use the request shapes in workflow.

Track, review, and deliver

Store the returned task_id and poll only that task through beatra.tasks.get. When the task ID is lost, use beatra.tasks.list to find candidates and verify the chosen task with tasks.get. Only when the original creation response is genuinely unknown may the byte-equivalent payload be replayed with the same client_request_id. Slow polling, connection, update, or authorization failures never create replacement paid work.

Cancel only when the user asks. If cancellation returns 409, continue tracking the original task. When the artifact is visible, review phone-size focal clarity, safe-area contrast, final canvas, crop resilience, and requested must-keep details. Deliver artifact links, observed dimensions, task ID, resolved model, and returned billing.net_charged_credits; offer one focused, unexecuted refinement only.

References by task

Updates and security

This Skill runs only the files installed with it and leaves them unchanged while it runs; it never downloads or executes other code. It talks to Beatra's official services over HTTPS, and a file you choose to upload goes only to the one-time upload address Beatra issues for it.

Once a day at most, the bundled client reads Beatra's public version list. When the store or tool you installed from has a newer version, it prints one line with the command that updates it there. Updating stays your choice, through that store or tool. See updates and security.