360 Secure SkillHub
360 安全技能中心

wedding-opening-film

核心功能:将新人提供的仪式事实转化为三个带标注的故事板关键帧(2K 16:9),再生成一个2-15秒的开场电影片段,不拼接。典型场景:摄影工作室或婚礼策划师制作婚礼开场片。输入:已有的视觉方向(情绪、调色板、仪式类型、语言)和三个命名故事板时刻。输出:三个关键帧图像,然后一个视频片段,每次付费前需确认。

beatra-ai
beatra-ai
热度 38.95
最近更新 2026-09-30 21:07:54
分类:创作依赖:MCP 服务 / 代码执行应用场景:图像生成 / 视频编辑 / 故事创作 / 工作流

安装技能

复制下面的话术发给你的 Agent,它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill: npx secure-skills-cli add wedding-opening-film
Metadata
name
wedding-opening-film
description
Turn couple-supplied ceremony facts into three wedding opening film storyboard keyframes, then one opening film. This wedding opening film studio lays out a couple storyboard and a ceremony opening clip from the facts you already have. Use it for photo-studio wedding openers, wedding-planner opening films, and couple storyboard keyframes.

Wedding Opening Film

Turn couple-supplied ceremony facts into three labeled storyboard keyframes, then one opening film. The first delivery is those three stills. The conversion is one 2–15s clip. Do not stitch.

Scope and adjacent routes

Use this Skill when a photo studio or wedding planner wants a wedding opening film from facts they already have: a couple storyboard, a ceremony opening clip, or a planner opener for the start of the event.

Route a product-launch opener to product-launch-opening-film. Route a script storyboard that is not a wedding opener to ai-storyboard-generator. Route generic video that is not this opener to beatra-ai-video-studio. Route a score that is not this film to music-generation-studio.

Collect the ceremony facts

Hard inputs are:

  • visual direction the studio already has (mood, palette, ceremony type, language);
  • three named storyboard moments for the opener.

Reuse already-known direction. Ask only for a missing hard input. Do not invent a name, date, or venue to finish a frame or a clip. A supplied photo is a later reference, not a source for missing identity, and it does not replace the three generated keyframes.

A longer ceremony brief is welcome. The first paid work is still three keyframes. The opening film waits until those stills are accepted.

Plan the free storyboard

Write a labeled three-keyframe sheet before any paid image: beat names, must-keep facts mapped to frames, mood, palette, language, and 16:9 2K canvas. Keep every supplied name, date, and venue exact. Leave a missing identity fact as a named gap. That sheet is the free visible result. Planning is not approval.

Safe defaults:

  • model: "auto" unless the user chose a live-eligible image model.
  • count: 1 per keyframe. Do not batch frames into one count.
  • canvas: {"type":"preset","tier":"2K","aspect":"16:9"}.

Confirm the three stills, then the film

Before any billable still, read the current beatra.models.list card for text_to_image. Show one current generate production card for the three keyframes and wait:

  1. Work — three wedding-opener storyboard keyframes, each a generate call (beatra.images.generate).
  2. Credits — the live text_to_image price just read, times three. Do not reuse a remembered number.
  3. Count — one paid call per named keyframe. Do not batch frames into one submit.
  4. Identity — one new opaque client_request_id per keyframe. A changed prompt, fact, file, model, or canvas mints a new ID.
  5. If we stop here — the labeled three-keyframe sheet remains usable.
  6. If the balance is insufficient — relay the official message and its top-up URL exactly. Translate the prose; keep the URL. Do not retry until the studio says they have topped up.

Submit each still once through bundled scripts/mcp_client.py. Keep at most two generate tasks in flight. Poll beatra.tasks.get. Deliver actual bytes, actual dimensions, and billing.net_charged_credits.

After those three stills are accepted, the opening film is a new pack. Read the current beatra.models.list card for text_to_video. Show a fresh six-field card for one beatra.videos.generate call and wait:

  1. Work — one wedding opening film, a generate call (beatra.videos.generate).
  2. Credits — the live text_to_video price just read. Do not reuse a remembered number.
  3. Count — one paid call for one clip. Do not batch stills into one submit. Do not stitch, concat, or interpolate the three stills.
  4. Identity — one new opaque client_request_id. A changed prompt, duration, model, or control mints a new ID.
  5. If we stop here — the three accepted stills remain usable.
  6. If the balance is insufficient — relay the official message and its top-up URL exactly. Do not retry until the studio says they have topped up.

Keep model: "auto" unless the user chose a live-eligible video model. Choose a supported integer duration in the 2–15s range; if they do not pick, use 8s. One clip. No stitch.

Do not treat acceptance of the stills as approval of the film.

An optional later clip that uses the three accepted stills as ordered image references is another new pack. Read {"capability":"reference_to_video"}, show a fresh six-field card for one beatra.videos.generate_from_references call, and wait. Do not reuse the stills card or the text-to-video card. Do not stitch.

Review, deliver, and recover

Review visible identity against the must-keep list. Report an invented name, date, or venue as drift. Report unread small type as unread. Do not promise exact seconds.

After a returned task_id, poll that task. If the create response is lost, search with beatra.tasks.list and verify with beatra.tasks.get before replay. Reuse an ID only with byte-identical arguments. Cancel only when the studio asks.

Execution

Invoke every remote Beatra operation only through this package's bundled scripts/mcp_client.py. Put the MCP tool name after call and send one JSON object on standard input.

python3 scripts/mcp_client.py call beatra.models.list
{"capability": "text_to_image"}
{"capability": "text_to_video"}
{"capability": "reference_to_video"}
printf '%s' '{"prompt":"Create approved storyboard keyframe 01 of 03 for this wedding opening film. Use only the confirmed ceremony facts. Do not invent a name, date, or venue.","model":"auto","count":1,"canvas":{"type":"preset","tier":"2K","aspect":"16:9"},"client_request_id":"opaque-wedding-still-01"}' | python3 scripts/mcp_client.py call beatra.images.generate
printf '%s' '{"prompt":"One wedding opening film from the approved three-keyframe storyboard. Keep confirmed ceremony facts. Do not invent a name, date, or venue.","model":"auto","duration":8,"client_request_id":"opaque-wedding-film-01"}' | python3 scripts/mcp_client.py call beatra.videos.generate
printf '%s' '{"references":[{"kind":"image","media":{"type":"artifact","artifact_id":"art-wedding-still-01"}},{"kind":"image","media":{"type":"artifact","artifact_id":"art-wedding-still-02"}},{"kind":"image","media":{"type":"artifact","artifact_id":"art-wedding-still-03"}}],"prompt":"One wedding opening film guided by these three approved storyboard stills. Keep confirmed ceremony facts. Do not invent a name, date, or venue.","model":"auto","duration":8,"client_request_id":"opaque-wedding-refs-01"}' | python3 scripts/mcp_client.py call beatra.videos.generate_from_references

Do not configure or call a host Beatra Connector, and do not use REST/OpenAPI as a fallback.

Account balance

When the user asks how many credits remain or whether a live estimate fits, call beatra.wallet.get. When they ask what was charged, call beatra.wallet.ledger. Both are read-only. Do not invent an account-balance or top-up tool. Do not make wallet.get a required step before every paid submit.

When a model card comes back carrying a top_up block, relay its tiers as the card lists them and in that order. Do not rank them, do not talk one down, and do not pick one for the user. Which tier suits them is their call, made on the wallet page with the whole list in front of them. Never quote a tier from memory.

References by task

Runtime and safe automatic updates

The bundled client silently checks for a newer release at most once every 24 hours per installation. When a newer version is available, it installs automatically without separate confirmation. It downloads only from the fixed official Beatra discovery and immutable CDN paths for this package, channel, and locale, verifies discovery data, archive, manifest, and every packaged file, and replaces only package-owned files.

Update checks, downloads, verification, replacement, rollback, and recovery fail open: the current installation remains usable and the original command continues. An update failure never authorizes retrying a paid image or video request. The setting persists for this installation. See automatic updates and safety.

python3 scripts/mcp_client.py update --auto off
python3 scripts/mcp_client.py update --auto on
python3 scripts/mcp_client.py update --check