Update to Voyager Video & 3D

{
  "description": "Cloud: generate short AI video clips with Voyager from a prompt, a start image or references, including saved Cast characters.",
  "included_files": [],
  "name": "voyager-cloud-video",
  "skill_md_contents": "---\nname: voyager-cloud-video\ndescription: \"Cloud: generate short AI video clips with Voyager from a prompt, a start image or references, including saved Cast characters.\"\n---\n\n# Short videos\n\nFollow `voyager-cloud` (read it with `load_skill` if it isn't loaded) for account,\nmodels, costs and jobs. This plugin\nmakes short clips one at a time. Editing, cutting several clips together, titles and\ntimelines are not available here; if the local Voyager plugin is installed, use it\nfor that.\n\n## Real product UI is never generated\n\nNever generate a real app's screens, a website or a real product's interface:\nmodels invent text and controls. Ask for a real screen recording or screenshot\nand use it as a reference, or show the product in an abstract way.\n\n## Generate a clip\n\n1. Choose the video model as `voyager-cloud` describes: exactly the model the\n   user named, otherwise the default (omit `model`). For a clip from a still the\n   recommended image-to-video model takes the start image. Use only durations,\n   aspect ratios, references and audio options the model supports, and pass a\n   start image or other references as asset IDs in `reference_ids` or in the\n   reference fields the model lists.\n2. For a clip from a still, make or import the still first and show it; animate\n   the still the user approved, and spend the video prompt on action and camera.\n3. Pass Cast IDs for every recurring character on every clip and retake.\n4. Call `generate_video` and poll\n   `get_job`. Video takes minutes: tell the user it is running and poll the same\n   job; never submit it again because it is slow.\n\n## Inspect and deliver\n\nLook at the result before presenting it: check the subject, motion, the\ncharacter's likeness and any requested words. A clip that drifted from the\napproved still or Cast look is a failed take; say what failed and offer one\nretake with a corrected prompt. Deliver with the download from `get_asset`.\n\n## Exclusions\n\nEvery generated clip risks the same defects. Exclude this base set unless the\nbrief wants one of them: on-screen text and signage, logos, watermarks, morphing\n(objects or bodies changing shape), extra limbs or objects, glow, particles and\nlens flares, and palette drift. Use a negative-prompt field only when the model's `list_models` settings\n  include one, passing it in `parameters` with the base set and any scene\n  exclusion as nouns. Otherwise describe the\n  wanted state (\"an empty, dry street\", \"plain unmarked storefronts\", \"colours\n  steady from start to end\"); a brief closing line of exclusions (\"No umbrellas,\n  no shop signs.\") is fine. Never describe the unwanted thing positively to limit\n  it (\"only two or three umbrellas\", \"one woman with a red umbrella\", drizzle),\n  and keep it out of the keyframe stills: what the prompt or a still shows, the\n  model renders.\n\n## The style key wins over photo references\n\nImage models follow a photo more readily than words: given a photo of a person\nand \"line art\", they tend to return a photo of the person. Once a style is set\n(a requested look such as line art or a storyboard sketch, or a style key image),\nhold it:\n\n- For a recurring character, pass Cast IDs on every\n  image/video estimate and run. The adapter adds its fixed description, refs\n  and model token mapping. Write shot action, composition, camera, look and\n  sound around the character's name; do not invent numbered Cast-ref tokens.\n- Pass a photo only for identity: the face, hair, wardrobe, a product's shape and\n  markings, a prop. Say which image does what, and that the photo does not set\n  the look: \"Use image 2 only for the woman's face, hair and coat. Draw the whole\n  frame as loose black line art on white, like image 1, not as a photograph.\"\n- Pass the style key image, when there is one, with every frame, and repeat its\n  words in every prompt. A shot under a film's scoped look (a photoreal\n  flashback) takes that look's image and words instead.\n- Before accepting a frame, compare it with the look that applies to it: the\n  same medium, line, fill, texture and palette. A frame that matches its scoped\n  look is not a failed take. A frame that drifted toward the photo is a\n  failed take, not a variant. Keep the same `cast` IDs and strengthen the style\n  direction on a correction. Do not drop identity refs or substitute text-only\n  identity. Report\n  likeness results separately from style inspection: a face match can still be\n  off-look, and an unavailable check is not a match.\n"
}