Update to Voyager Video & 3D

{
  "description": "Cloud: Voyager account, credits, estimates, references, jobs, results, brand kits and Cast for hosted generation. Read first for any Voyager image, video, audio or 3D request in ChatGPT or Codex.",
  "included_files": [],
  "name": "voyager-cloud",
  "skill_md_contents": "---\nname: voyager-cloud\ndescription: \"Cloud: Voyager account, credits, estimates, references, jobs, results, brand kits and Cast for hosted generation. Read first for any Voyager image, video, audio or 3D request in ChatGPT or Codex.\"\n---\n\n# Voyager in the cloud\n\nThis plugin generates media on Voyager's servers with the user's existing Voyager\naccount, selected workspace and credits. It has no access to the user's computer,\neditors or project folders. If the local Voyager plugin is also installed and the\nwork involves files on this computer, a timeline or an editor, use the local plugin\nfor that part.\n\nRead the family skill before generating: `voyager-cloud-image`, `voyager-cloud-video`,\n`voyager-cloud-audio` or `voyager-cloud-3d`. When a skill isn't already loaded,\nread it with `load_skill` and its name; a skill's reference files load the same\nway with `file` set to the reference's path.\n\n## Account and credits\n\nA plain request needs no account check first: a generation that the workspace\ncan't pay for is refused with the reason and a link. Call `account_status` when\nthe user asks about their plan or credits, or after a billing failure: it reports\nthe selected workspace, whether generation is included for this user, seat state\nand the credit balance.\n\nWhen credits, plan or seat state don't allow a generation, the tool's error (or an\nestimate's reason) says why in one sentence and links the workspace's Voyager plan and\ncredit details page. Show that sentence and link as given,\nthen ask how to continue. Never retry on another model to get around it without\nasking. On `rate_limited`, wait the time the error gives and retry once.\n\n## Models and settings\n\nModels, supported settings, durations and limits change often: never rely on a\nmodel name, price or limit from memory. Spend at most one discovery call before\nthe estimate.\n\n- **The user named no model:** generate with the default: omit `model` and the\n  generation tool uses it (the result names it). When the request clearly fits a\n  recommendation better (a 20-second shot, many references, faces and dialogue,\n  clean text in an image), call `list_models` with the `operation` once, use the\n  better fit and say why in a few words. With the first result of its kind in the\n  conversation, add one line of alternatives from the recommended list, each with\n  its reason (for video, for example: \"fast and cheap: H3 Max Turbo · atmosphere\n  and longer shots: Seedance 2.5 · cinematic realism: FLUX 3 Max\"). Don't ask the user\n  to choose a model first.\n- **Common asks switch the model without a question:** cheaper or faster video →\n  H3 Max Turbo; longer than 15 seconds or many references → Seedance 2.5;\n  cinematic realism → FLUX 3 Max; faces and spoken dialogue → Veo 3.1; clean text,\n  a logo or a transparent background in an image → GPT Image 2.5 Flare. Say why\n  in a few words.\n- Ask at most one question, and only when something you can't sensibly default\n  is missing (a reference the user mentioned but didn't attach, for example).\n  Don't batch questions.\n- **The user named a model:** use exactly that model, and keep using it for\n  follow-ups on the same work until the user names another. Find it with one\n  `list_models` call with `query` set to the name. When nothing matches, or the\n  name fits more than one entry (a family name, a missing version or tier), say\n  so, show the closest entries, and ask once which to use. Never pick a near\n  match, a newer or cheaper version, or another provider in its place.\n- Never page through the whole catalog and never estimate several models to\n  compare them. Never invent a model name or ID, and never switch models because\n  another one's settings look simpler. Call `list_models` with `model` only when you need settings\n  beyond prompt, duration, aspect ratio and audio.\n- Name the model with the result. If Voyager refuses the model, or the job\n  reports a different one, stop and ask before going further.\n\nPass only settings that `list_models` reports for that model: the named tool\ninputs where they exist, and any other setting in `parameters` under the field\nname the model's listing uses, with every media URL imported with\n`import_reference` and replaced by its asset ID.\nA\nrequested setting the model does not support (a duration, aspect ratio or audio\noption) is also a question for the user, not a reason to switch models.\n\n## Generate\n\n1. For a plain request, call the generation tool directly; everyone expects media\n   to cost credits. Call `estimate_generation` first only when the user asks what\n   it will cost, gives a budget, or the job is unusually large (several clips, a\n   long video, many images): then show the model and estimated credits and wait\n   for their go-ahead. An estimate creates no hold and is not a final price.\n2. The generation tool waits up to 20 seconds and returns the job, finished or\n   still running.\n3. While the job runs, call `get_job` with its ID and `wait_seconds` (up to 25)\n   again and again until it succeeds, fails or is cancelled, then show its\n   outputs with the model used and the credits charged. Don't end your turn\n   while a job you started is still running.\n\nGive every generation call a fresh `client_key`. If a generation call's response\nis lost, repeat it with the same parameters and the same `client_key`: Voyager\nreturns the original job instead of paying twice. Never resubmit with a new key\nbecause a call timed out or a job looks slow: poll the same job with `get_job`. A job in `unknown` state is still being reconciled:\nwhile `get_job` still suggests a poll delay, keep polling the same job. When it\nsuggests none, stop polling, tell the user the outcome is uncertain, and check\nthe same job again later. Never start a second paid attempt for it. `list_jobs` finds a job when\nthe ID is lost. `cancel_job` is best effort: it does not promise a refund, and\ncredits already spent stay spent. Make no automatic paid retries and no model\nswitches without the user's agreement; ask before any new paid attempt.\n\n## References and results\n\nReferences come only from files the user attached in this conversation, HTTPS\nURLs they gave you, or a local file the user named when you run on their\ncomputer (a CLI such as Claude Code or Codex). Call `import_reference` for each,\nand pass the returned asset IDs to generation tools. For a local file, call\n`create_upload` with its name, run the `curl` command it returns with the file's\npath, then call `import_reference` with the returned `upload_id`. Never guess\npaths on the user's computer, and never look through the user's other files.\n\nImported references are kept for 7 days and generated results are downloadable\nfor 30 days. Use `get_asset` for a fresh download link; links expire after a few\nminutes, so fetch one when the user needs it instead of pasting an old one.\n`delete_asset` deletes an owned asset on request. Give results meaningful\nfilenames in your reply.\n\n## Brand kits\n\n`list_brand_kits` and `get_brand_kit` read the workspace's brand kits: palette,\ntypefaces and logo asset IDs. Use them when the user names a brand or asks for\non-brand work: put the palette and type into the prompt, and pass a logo asset ID\nas an explicit reference only when the logo should appear. Brand kits are\nread-only here; edits happen in Voyager.\n\n## Cast\n\nCast cards are named, saved characters (or recurring products and props) with a\nfixed description and reference images. They persist in the workspace until\ndeleted and are shared by the user's ChatGPT and Codex connections to the\nworkspace. They are separate from Cast in the Voyager app's local projects.\n\n- Before making a character, `list_cast` and `get_cast` to reuse a fitting card.\n- To make one from scratch, `generate_cast` creates its reference images through\n  the normal billed image flow.\n- To save attached images as a card, import them, then `save_cast` with a short\n  fixed description (face, hair, build, wardrobe; never the medium such as\n  \"cartoon\") and the asset IDs. Saving existing images is not billed.\n- To change a card, `save_cast` with its ID and the `expected_version` you read; on\n  a version conflict, read it again and reapply the user's change.\n  `add_cast_reference` and `remove_cast_reference` change its images;\n  `delete_cast` deletes the card, only when the user asks.\n- Pass Cast IDs to `generate_image` or `generate_video` on every shot with that\n  character, including retakes. Write the action, camera and look around the\n  character's name.\n\n## Result cards\n\nResults show as cards with previews and downloads where the host supports them;\notherwise give a short text summary with the job ID, filenames and download links\nfrom `get_asset`. Don't autoplay or describe a result you haven't received.\n"
}