Use when generating or editing video with Luma Ray models on Scenario via MCP: cinematic text-to-video, image-to-video from a start frame, start plus end frame anchors, seamless looping clips, HDR output, restyling footage while preserving motion with edit strengths and face, pose, depth, or trajectory controls, reframing to another aspect ratio by outpainting, or budget prompt edits on real footage. Keywords: Luma, Dream Machine, Ray 3.2, Ray 3, Modify Video, Reframe, T2V, I2V, V2V, loop, HDR.
インストールする前に、エージェントに与えられる指示の中身を確認できます。
Luma's video line on Scenario is four sibling models, one per mode: Ray 3.2 generates from text or frame anchors, Ray 3.2 Edit restyles footage under structural controls, Ray 3.2 Reframe outpaints to a new aspect ratio, and Modify Video makes budget prompt edits. Route by member before tuning anything; Luma's image models belong to the scenario-luma-image skill. Discover the live set with search and treat model_schema_get as the contract.
Connection and the core loop: see the scenario skill in this repo; model-agnostic video work: the scenario-video skill. If a sibling skill named here is missing from your available skills, ask the user to install it (npx skills add scenario-labs/skills --skill <name>); unattended, proceed from tool schemas and flag the gap.
Each mode is its own model (names from the live schemas, caps as of authoring time):
| Member | Inputs | Behavior |
|---|---|---|
| Ray 3.2 | prompt (+ startFrame, endFrame) | text or image to video; endFrame is only valid alongside startFrame |
| Ray 3.2 Edit | video + prompt | restyles while keeping motion and timing; editStrength plus controls |
| Ray 3.2 Reframe | video + prompt + aspectRatio | fits a new ratio by outpainting, never cropping; prompt fills the canvas |
| Modify Video | video (+ prompt, firstFrame) | budget prompt edits via a mode ladder; source up to 30 seconds, 100 MB |
On Ray 3.2 the options veto one another. At authoring time duration was a unit string, "5s" or "10s", never a number; "10s" refused loop, hdr, startFrame, and endFrame. hdr required 720p or 1080p from the 540p, 720p, 1080p resolution list; six aspectRatio values ran 9:16 through 21:9. Edit and Reframe share that resolution list; Modify Video has no resolution parameter. duration, resolution, and hdr each move the price (1080p ran several times 720p's cost and wait), so dry_run the option set before a batch.
Ray 3.2 Edit and Modify Video restyle through the same nine-value ladder, adhere_1 through reimagine_3 (adhere stays close, flex restyles but keeps recognizable elements, reimagine transforms), and share little else. Edit names the ladder editStrength; Modify names it mode. Edit takes a guide image as startFrame; Modify's firstFrame is an edited copy of the source's own first frame, not an arbitrary style image. Edit alone offers keyframes (up to 64 image-and-index anchors, mutually exclusive with the single startFrame guide), face, pose, depth, normals, and trajectory conditioning toggles, resolution choice, and hdr; or pass autoControls: true instead of a manual editStrength. The price gap ran near two orders of magnitude at authoring time, so dry_run the same clip on both and pay for Edit's controls only when the edit needs them.
Ray rewards natural-language prompts of roughly 50 to 300 words built around motion: a subject mid-action, one named camera move (slow push-in, handheld follow), and one physical consequence of the action (droplets scattering, dust rising). Concrete lighting language lands directly. With loop, hold motion intensity constant (steady rain, drifting steam) so the cycle closes cleanly. On Reframe, prompt only the edges: describe what the new canvas should contain and leave the original subject alone.
search with target="models", query="luma", public=true. Prefer the newest non-deprecated generation, e.g. model_luma-ray-3-2 and its Edit and Reframe siblings (live hits at authoring time: re-discover each session).model_schema_get with the generator id: options and their vetoes before anything else.upload_asset the product still (see the scenario skill) to get an asset id.model_run with that model_id, dry_run=true, and the exact parameters={"prompt": "A crystal perfume bottle catches soft window light as a single drop arcs off the stopper, slow push-in, product commercial style.", "startFrame": "asset_a", "duration": "5s", "resolution": "1080p", "aspectRatio": "16:9"}; re-estimate after any option change.model_run with wait=false, then jobs_wait with the returned job id, re-called with pending_job_ids on timeout, never a second model_run.model_schema_get the Reframe sibling, then model_run with parameters={"video": "<generated asset id>", "aspectRatio": "9:16", "prompt": "continue the marble counter downward and the softly lit wall upward", "resolution": "1080p"}, same wait discipline.asset_display both and inspect the outpainted edges before delivery.duration: 5 or "5": Ray 3.2 takes the unit string, "5s" or "10s"."10s" excludes all three; drop to "5s".firstFrame and mode, not startFrame and editStrength.editStrength with autoControls, or startFrame with keyframes: each pair is mutually exclusive on Edit.まだレビューはありません。使ってみた感想をお寄せください。
概要と使いどころ
Handle review comments received on GitHub for the current branch's PR or a given PR number or URL, by triaging each unresolved thread, fixing what is warranted, pushing, and replying in each thread, when asked to address, fix or answer PR comments. Works in any Scenario repository.
日本語の概要は準備中です。原文の説明を表示しています。
Generate one short Conventional Commits squash message describing only the net diff between the PR's base and head, set its header as the PR title, validate it with the repository's commitlint and both copy its description to the clipboard and display it, when asked for a squash message. Works in any Scenario repository, whatever its default branch.
日本語の概要は準備中です。原文の説明を表示しています。
Refresh the current pull request description from its net changes, following the repository's PR template, when asked to update or write the PR summary. Works in any Scenario repository, whatever its default branch.
日本語の概要は準備中です。原文の説明を表示しています。
Use when connecting an AI agent to Scenario (scenario.com) through MCP, or when a task involves generating images, video, 3D, audio, sprites, textures, or game assets. Also when picking a Scenario model, running a LoRA, refining a generation prompt, uploading reference images, waiting on generation jobs, checking credits or quota, hitting Scenario auth, scope, or Forbidden errors, or setting up mcp.scenario.com in Claude Code, Cursor, VSCode, or another agent.
日本語の概要は準備中です。原文の説明を表示しています。
Use when generating or handling 3D assets through the Scenario MCP server, including text-to-3D or image-to-3D meshes, GPT-6 Astra 3D or Claude Opus 5.5 3D, GLB, FBX, OBJ, STL, or VOX files, PBR-textured or game-ready models, voxel models, multi-view reconstruction, retexture, remesh, UV unwrap, auto-rigging a biped or quadruped, or retargeting an animation, previewing a mesh in the inline 3D viewer, or downloading a model for import into Unity, Unreal, Godot, or Blender.
日本語の概要は準備中です。原文の説明を表示しています。
Use when generating navigable 3D worlds or Gaussian splat scenes on Scenario via MCP: text-to-world, image-to-world, a 360 panorama or skybox turned into a walkable splat, multi-view photos or a video walkthrough reconstructed as a splat, or one object photo to splat. Keywords: World Labs Marble 1.1, 1.1 Plus, 1.0 Draft, Hunyuan World, HY World, image-to-skybox, skybox-to-splat, multiview-to-splat, TripoSplat, gaussian splatting, 3DGS, spz, world generation, explorable scene, VR.
日本語の概要は準備中です。原文の説明を表示しています。