Posed RGBD generation

atlasGenerate generates posed RGB views at target cameras from one or more posed RGBD context frames. Each context frame supplies an image, depth buffer, and camera so the model can preserve the submitted scene geometry.

Account availability: call POST /api/v2/tasks:atlasGenerate only when it appears in the API reference for the selected account.

Prepare RGBD context

Every contextFrames entry requires an imageAsset, depth, and camera. Keep those three values aligned to the same view and coordinate system. Target cameras determine how many frames are generated and their order in the response.

Targets must use the model's 1280 × 720 camera grid. Omit model to let Marble route the request based on its context and target counts. Pin a model only when you have evaluated that checkpoint for your workflow.

Prompt enhancement is enabled by default. Preserve promptUsed from the response so the actual model conditioning remains auditable.

Generation uses the selected model's sampling defaults. The current distilled models use eight denoising steps with guidance baked into their weights. Omit modelParameters.numSteps to use the model's sampling recipe.

Request output depth

Set returnDepth: true to reconstruct depth for the generated views and attach it to each returned frame. This will fail when the camera origins between context and targets are all identical, as the depth scale is ambiguous in this scenario.

Example completed output

After the operation reports done: true, its task-specific result is in operation.response. This example requested output depth:

{  "frames": [    {      "imageAsset": {        "assetId": "asset_generated_view_0",        "url": "https://example.com/generated-view-0.png"      },      "camera": {        "extrinsics": {          "position": [0.5, 1.6, -1.2],          "quaternion": [0, 0, 0, 1],          "coordinateSystem": "rub"        },        "intrinsics": {          "width": 1280,          "height": 720,          "fx": 900,          "fy": 900,          "cx": 640,          "cy": 360        }      },      "depth": {        "depthAsset": {          "assetId": "asset_generated_depth_0",          "url": "https://example.com/generated-view-0.exr"        },        "confidenceAsset": {          "assetId": "asset_generated_confidence_0",          "url": "https://example.com/confidence-0.png"        }      }    }  ],  "promptUsed": "A quiet stone courtyard at blue hour",  "requestId": "request_example"}

When returnDepth is false, each frame contains only its generated imageAsset and camera. Open the atlasGenerate reference for the live sequence limits, model choices, and complete schemas.