Mascot Scenes for Banners and Social Cards

Place your mascot in an illustrated, fully vector scene for banners, empty states and social cards.

A scene is an illustration with your mascot in it: a cozy desk for an empty inbox, a launch pad for a release banner. The scene agent draws the world around the character in the same vector language, places your actual rig in it, and returns both an SVG and a PNG. Because the mascot is your rig and not a redraw, it is on-model every time.

A scene costs 50 credits ($0.50 at plan rates), held when the job starts and refunded if it fails. See Credits & limits.

Create a scene#

POST/mascots/{id}/scenes

Name	Type	Description
`prompt`required	string	The setting, mood and purpose, 3–1000 characters. Say where copy will go if you need space, e.g. "leave the left half calm for a headline".
`size`	string	`square`, `wide` (default), `banner` or `portrait`.
`animation`	string	Pose the mascot from this animation: an id, a name or a preset id.
`t`	number	Seconds into `animation` to take the pose from. Default 0.4.
`max_cost_usd`	number	Abort the job if model spend passes this (it can only lower the built-in ceiling).

curl -X POST https://lumorig.com/api/v1/mascots/msc_…/scenes -H "Authorization: Bearer $LUMORIG_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "prompt": "cozy desk at night for an empty inbox, leave the left half calm for copy",
        "size": "banner", "animation": "sleep" }'
# 202 → { "id": "job_…", "type": "scene", "status": "queued", … }

Sizes#

size	Canvas	Good for
`square`	1024 × 1024	Social posts, avatars, app-store tiles.
`wide`	1600 × 900	Hero banners and slides. The default.
`banner`	1600 × 600	Email headers, page headers, empty states across a wide column.
`portrait`	900 × 1200	Stories, mobile onboarding screens, posters.

The SVG uses these dimensions as its coordinate space, so it scales to any display size without loss.

Posing the mascot#

Without animation the mascot stands in its rest pose. With one, the agent receives the pose at time t of that clip, so a celebrate scene shows arms up and a sleep scene shows closed eyes. Pick t at the most readable moment of the motion; scrub it first with the preview endpoint (?animation=celebrate&t=0.6), which is free.

Results#

When the job succeeds, job.result holds two files and a short description:

Field	What
`svg`	The full scene as vector: file object with id, url, mime image/svg+xml, bytes.
`png`	The same scene rasterised: file object with mime image/png.
`summary`	One or two sentences on what the agent drew.

Download either with GET /files/{id} and your API key. The CLI saves both as <id>-scene.svg and .png in the -o directory; the local MCP server does the same with save_to.