360 Secure SkillHub
360 安全技能中心

amazon-main-image-motion

将卖家已通过商品图规则的某一张白底主图作为严格首帧,生成一条用于 Amazon 主图视频位的短视频片段(每张图一条、不拼接)。流程含免费镜头计划、模型能力校验、六要素制作卡片确认、单次付费提交、任务轮询与结果复核,全部远程操作走包内 MCP 客户端。

beatra-ai
beatra-ai
热度 60.25
最近更新 2026-10-12 18:13:03
分类:创作依赖:代码执行 / MCP 服务 / 第三方 API / 读写文件应用场景:图像生成 / 视频编辑 / 品牌素材 / 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 amazon-main-image-motion
Metadata
name
amazon-main-image-motion
description
Turn one approved white-background main image into one short clip for the listing main video slot. The approved pack shot is the first frame, so the clip opens on the product and the white background you already cleared, and from there it turns, a light sweeps across it, or the camera eases in. Use it for Amazon main image video, main image motion, product photo animation, and white background product video work that stays one photo one clip.

Amazon Main Image Motion

Turn one approved white-background main image into one short clip for the main video slot. Keep the pack shot the seller already cleared. One photo makes one clip. Do not stitch.

Scope and adjacent routes

Use this Skill when a seller has a main image that already passes the listing's image rules and wants that exact frame to move.

Route a spoken product demo to product-video-studio. Route a still that is not an approved listing main image to image-to-motion. Route the still main image set itself to amazon-main-image-set. Route a set of still variants for the A+ module to amazon-a-plus-module-pack. Route per-step assembly photos to assembly-one-step-clip. Do not reshoot, recompose, or re-background the main image here, and do not write the listing copy. Do not look up public posts, reviews, or competitor listings.

Inputs and defaults

The hard inputs are:

  • one accessible main image the host Agent can inspect;
  • what the seller wants to move, in their own words.

Everything else is optional. Reuse SKU name, language, marketplace, and must-keeps already in the conversation. Ask only for a missing hard input. If the label text in the image is too small or too blurred to transcribe, stop and ask for a cleaner file rather than guessing the words.

Do not invent a claim, certification, size, or count that the seller did not state. The image is not a source for unstated facts.

Product motion stays inside this vocabulary. Pick one. Subtle beats dramatic — a large requested movement distorts the product and its label.

  • Always available, because they reveal nothing the image does not already show: turntable rotation, light sweep, gentle push-in.
  • Only when the seller supplies what is inside — an interior photo or a stated contents list: lid lift, unfold, exploded view. Without that, the model would invent the interior. Ask for it or offer a turntable instead.

Default to one beatra.videos.animate call, model: "auto", the main image as the strict first frame, no driving_audio, and no aspect_ratio, which leaves the ratio to the card: a declared model default, a source-derived ratio, or fixed model behaviour. Read the chosen model's live card. Unless the card promises a source-derived ratio, compare its output ratio with the image's, and if they differ, tell the seller the pack shot will be re-framed before showing the production card. Choose a supported integer duration in the 2–15s range with the seller; if they do not pick, use 5s rather than the 15s ceiling.

Golden path

  1. Inspect the image. Record MIME type, width, height, aspect ratio, byte size, and whether it has an alpha channel. Transcribe every legible string on the pack — brand, variant, size, count, claim — verbatim. For a local file, upload only through the bundled client after inspection. Never pass a local path to a remote tool.

  2. Write a free shot plan the seller can check before anything is billable:

    • the opening state at 0.0s, naming what is not yet present so a reveal has somewhere to come from;
    • the one motion from the vocabulary above, with its direction and speed;
    • the closing hold;
    • every transcribed string, quoted, marked to stay pixel-identical;
    • what the plan locks: background white, product colour, and every quoted string. It does not lock the crop, because the card's output ratio can re-frame the shot. The plan is what the delivered clip is reviewed against; generated frames can still drift, and drift is reported rather than glossed.

    That plan is the free visible result. Planning is not approval.

  3. Call beatra.models.list with {"capability":"image_to_video"}. Admit the image against a current card that accepts a first-frame image without driving audio. Compare MIME, dimensions, bytes, and the chosen duration with advertised constraints. If any required media fact is unavailable or incompatible, stop before video.

  4. Show one current production card and wait before any video client_request_id or beatra.videos.animate call:

    1. Work — one main image clip from one named image, a generate call (beatra.videos.animate).
    2. Credits — the live image_to_video price just read. Do not reuse a remembered number.
    3. Count — one paid call per image. Two SKUs are two cards. A second take of the same image is a second charge; say the multiplication out loud before the seller approves.
    4. Identity — one new opaque client_request_id per image. A changed prompt, image, duration, model, or control mints a new ID.
    5. If we stop here — the shot plan 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 seller says they have topped up.

    Do not submit until the seller confirms they have topped up or already have enough credits.

  5. Submit beatra.videos.animate exactly once per approved image through bundled scripts/mcp_client.py. Do not configure a host Beatra Connector. Do not use REST/OpenAPI as a fallback. Poll the video task with beatra.tasks.get until terminal.

  6. Deliver the clip with actual dimensions, duration, usage, and billing.net_charged_credits. Review it against the plan: did the named motion fire, did the background stay white, did every quoted string stay readable at full size. Report drift as drift. Never invent a stitch, concat, or editor tool.

Decisions that require confirmation

Shot planning and live price quotes are free. They are not approval. Video animate needs the six-field card above. Re-running the same image because the seller wants a different look is new paid work with a new card and a new ID. File access is not consent to reuse the image outside this clip.

Recovery

The paid image has one frozen payload and one ID. Recover a lost create response only with that identical payload. Recover a lost task ID through beatra.tasks.list and beatra.tasks.get. Call beatra.tasks.cancel only when the seller asks to cancel; on 409 keep polling. On insufficient_balance, keep the top-up URL inside the balance error exact and retry the same frozen ID only after the seller says they have topped up. If a create returns no task_id, do not poll; reconcile before minting a new ID. A changed image, prompt, duration, or model is new work and a new card.

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": "image_to_video"}

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

If the tool returns insufficient credits: give the user the tool message in full. Do not change the URL. You may translate the other words. Do not replace it with your own top-up copy. Point the user to pay using that message.

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 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