RS 3D Pose Masks (Experimental)
Experimental 3D mannequin composition editor. Rotate joints and move people/camera; each character number creates a 2D region mask. Full silhouettes overlap; visible silhouettes remove occluded areas. Canvas overflow is clipped. Editing and masks need no extra weights. A composition guide, not exact pose enforcement. Connect only building_pipe before Finalize. The sampler automatically loads and applies matching Anima 2B/Krea2/MOSS pose adapters; no MODEL, VAE or patch-loader connection is needed here. Auto on other architectures uses masks only and reports the reason. New scenes use a native MHR skinned body. Anima 2B pose following has been observed with a separately trained adapter; exact joints and contacts are not guaranteed. Legacy ellipsoid scenes are preserved. Masks alone do not enforce joint angles. Scene data is stored in the workflow. Connect anywhere before Finalize.
- building_pipe
- building_pipe
- mask_preview
- full_masks
- visible_masks
- scene_info
- pose_image
- scene_json
Inputs (6)
| Name | Type | Default | Description |
|---|---|---|---|
| scene_json | STRING | {"version": 2, "width": 768, "height": 1024, "mask_mode": "full", "camera": {"position": [0, 1.15, 4.1], "target": [0, 0.9, 0], "fov": 38}, "people": [{"id": 1, "position": [-0.4, 0, 0], "rotation": [0, 0, 0], "scale": 1.0, "bulk": 1.0, "joints": {"left_shoulder": [5.701623648928897, -30.796652039695918, -20.70351733941452], "left_elbow": [0.9020204954093949, -3.030985637358303, -33.13930901996803], "right_shoulder": [5.701623648928899, -30.796652039695918, -20.70351733941453], "right_elbow": [0.9020203333155552, -3.0309861422158897, -33.13930977533524]}}, {"id": 2, "position": [0.4, 0, 0], "rotation": [0, 0, 0], "scale": 1.0, "bulk": 1.0, "joints": {"left_shoulder": [5.701623648928897, -30.796652039695918, -20.70351733941452], "left_elbow": [0.9020204954093949, -3.030985637358303, -33.13930901996803], "right_shoulder": [5.701623648928899, -30.796652039695918, -20.70351733941453], "right_elbow": [0.9020203333155552, -3.0309861422158897, -33.13930977533524]}}], "outline": true} | Open 3D editor, arrange mannequins, then Apply. Character numbers map to prompt slots. The scene is saved in the workflow. |
| building_pipeopt | RS_MC_BUILD_PIPE,RS_MC_CHARACTER_PIPE,NAI_CONTEXT | — | |
| pose_controlopt | COMBO | auto | Auto applies a matching pose adapter inside the sampler. Other model families use masks only. / 자동: 샘플러에서 모델에 맞는 포즈 어댑터 적용. 미지원 모델은 마스크만 사용. |
| pose_strengthopt | FLOAT | 1.000–2 | Pose guidance strength. 0 uses masks only and downloads nothing. / 자세 유도 강도. 0은 마스크만 사용. |
| pose_startopt | FLOAT | 0.000–1 | Start fraction of denoising. / 자세 유도 시작 구간. |
| pose_endopt | FLOAT | 1.000–1 | End fraction; must follow start. / 자세 유도 종료 구간. |
Outputs (7)
| Name | Type | Description |
|---|---|---|
| building_pipe | RS_MC_BUILD_PIPE | — |
| mask_preview | IMAGE | — |
| full_masks | MASK | — |
| visible_masks | MASK | — |
| scene_info | STRING | — |
| pose_image | IMAGE | — |
| scene_json | STRING | — |