apartment-kitchen-tour

Apartment Kitchen Tour

Turn one kitchen photo the listing already uses into one short clip for the listing page or a short-video slot. One photo makes one clip. Do not stitch a whole-home walkthrough.

Scope and adjacent routes

Use this Skill when an agent or host has a kitchen photo and wants that exact frame to move.

Route a set of rooms for a whole-home pack to listing-room-video-pack. Route a spoken welcome from a host photo to airbnb-welcome-avatar. Route a still that is not a listing room to image-to-motion. Do not shoot, retouch, virtually stage, or re-light the photo here, and do not write the listing copy or estimate a price. Do not look up public posts, reviews, or competing listings.

Inputs and defaults

The hard inputs are:

• one accessible kitchen photo the host Agent can inspect;
• what the agent or host wants to move, in their own words.

Everything else is optional. Reuse unit name, language, market, and must-keeps already in the conversation. Ask only for a missing hard input. If text in the photo — an appliance brand, a unit number, a price card — is too small or too blurred to transcribe, stop and ask for a cleaner file rather than guessing the words.

Do not invent an appliance, a fixture, a measurement, or an amenity the photo does not show and the agent did not state. The photo is not a source for unstated facts.

On-screen motion stays inside this vocabulary. Pick one. Subtle beats dramatic — a large requested movement warps counters and cabinet lines.

• Always available, because they reveal nothing the photo does not already show: slow push-in, light sweep across the counter, and parallax drift between depth layers so long as it stays small enough not to open a gap behind a foreground object.
• Only when the agent supplies what is inside — an interior photo of that cabinet, drawer, or appliance: opening a cabinet, pulling a drawer, opening the fridge. Without it the model invents the contents, and a renter or buyer reads storage and appliances off that clip. Ask for the interior photo or offer a push-in instead.

Default to one beatra.videos.animate call, model: "auto", the photo 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 photo's, and if they differ, tell the agent the framing will change before showing the production card. Choose a supported integer duration in the 2–15s range with the agent; if they do not pick, use 5s rather than the 15s ceiling.

Golden path

1. Inspect the photo. Record MIME type, width, height, aspect ratio, byte size, and whether it has an alpha channel. Transcribe every legible string in frame — appliance brand, unit number, price card — 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 agent 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 camera move from the vocabulary above, with direction and speed;
   • the closing hold;
   • every transcribed string, quoted, marked to stay unchanged;
   • the negatives the prompt will carry: no film grain, no vignette, no lens flare, no watermark, no extra text, and no element the photo does not already contain.

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

3. Call beatra.models.list with {"capability":"image_to_video"}. Admit the photo 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 kitchen clip from one named photo, 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 photo. Three rooms are three cards. A second take of the same photo is a second charge; say the multiplication out loud before the agent approves.
   4. Identity — one new opaque client_request_id per photo. A changed prompt, photo, 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 (https://console.beatra.ai/wallet?intent=buy). Translate the prose; keep the URL. Do not retry until the agent says they have topped up. Do not recommend ¥198.

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

5. Submit beatra.videos.animate exactly once per approved photo 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 move fire, did the room stay as photographed, 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 photo because the agent wants a different look is new paid work with a new card and a new ID. File access is not consent to reuse the photo outside this clip.

Recovery

The paid photo 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 agent asks to cancel; on 409 keep polling. On insufficient_balance, keep the top-up URL exact and retry the same frozen ID only after the agent says they have topped up. If a create returns no task_id, do not poll; reconcile before minting a new ID. A changed photo, 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.

References by task

• For shot plans, payloads, confirmation, polling, recovery, and review, read Kitchen tour workflow.
• For authorization and the non-billable registration step, read installation and authentication and installation registration.
• For shared task, billing, and connection details, read tasks and results, billing, errors, and recovery, and Bundled MCP Client diagnostics.
• For update guarantees and controls, read automatic updates and safety. For removal, read uninstall and disconnect.

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