Media generation

Image generation is synchronous: the response carries the finished image. Pick an engine or omit it — the default is theo-ultra on paid plans (Free-plan calls fall back to theo-creative and the response says so):

curl https://www.opencharts.com/api/v1/images \
  -H "Authorization: Bearer $OPENCHARTS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "prompt": "Isometric pixel-art data center, neon accents", "aspectRatio": "16:9" }'

POST /images/moonlight renders on Theo Moonlight, the first-party image model OpenCharts serves on its own hardware. It is the endpoint to point a benchmark or an arena at: synchronous, seedable, and with only a prompt it renders at the model's published defaults (1024x1024, 8 steps, guidance 1). The response carries the PNG bytes and the seed that was used; nothing is stored for you.

curl https://www.opencharts.com/api/v1/images/moonlight \
  -H "Authorization: Bearer $OPENCHARTS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "prompt": "A lighthouse keeper reading by lamplight, rain on the window, oil painting", "seed": 42 }'

Video renders return a 202 job envelope immediately:

Completed jobs return two versions, each with audio, cover art, and a Theo-branded engine label.

Build a full Code Canvas project (site or app) from a prompt; the finished project opens in the OpenCharts editor:

POST /real-world/spaces turns photos of ONE room into an imagined, walkable 3D room on Theo Real World. Send up to eight photos taken from one spot, turning between shots so each overlaps the last (imageUrls), or one short video of the room (videoUrl), and Theo works out where each photo was taken, builds the room, and imagines every wall, floor and corner the photos did not show. The job answers in two tiers on the same space_ job: a walkable quick room lands on the poll while the job is still processing (about a minute and a half), and the full room at real scale (with metricScale, a 360 panorama and a collider mesh) completes it a few minutes later. One photo alone (imageUrl) builds a room too; on a deployment without the room engine it becomes a measured, imagined, walkable space instead, with the 360 preview and the room's dimensions first.

curl https://www.opencharts.com/api/v1/real-world/spaces \
  -H "Authorization: Bearer $OPENCHARTS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "imageUrls": ["https://example.com/photos/living-room-1.jpg", "https://example.com/photos/living-room-2.jpg", "https://example.com/photos/living-room-3.jpg"], "quality": "full" }'

{
  "stage": "space",
  "imagined": true,
  "source": "photos",
  "photoCount": 3,
  "panoramaUrl": "https://www.opencharts.com/api/canvas-video/...",
  "spaceUrl": "https://www.opencharts.com/api/canvas-video/...",
  "spaceFullUrl": "https://www.opencharts.com/api/canvas-video/...",
  "format": "spz",
  "metricScale": 1.38,
  "floorOffsetM": 1.351,
  "colliderUrl": "https://www.opencharts.com/api/canvas-video/...",
  "depthUrl": null,
  "posterUrl": "https://...",
  "gaussians": null,
  "dimensions": null,
  "engine": "Theo Real World 1.1"
}

Register a webhook for space.completed / space.failed to be told when the job ends instead of polling. A completed job may stay at stage: preview when only the preview was asked for, or when the walkable tier did not finish; the result's notes say so. The full walkthrough, from the key to rendering the splat, is the Real World spaces guide.

Multi-step web research with citations, returned as a markdown report plus sources:

Poll research like every other job at GET /jobs/{id} using the returned res_… id. The legacy GET /research/{jobId} path stays for older integrations: it accepts both id forms and returns its original (pre-jobs) response shape.

curl https://www.opencharts.com/api/v1/research \
  -H "Authorization: Bearer $OPENCHARTS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "query": "State of AI-native documentation tooling in 2026" }'