Solid Render

Overview

The camera that converts a Solid SDF volume into a pixel-art asset. It uses orthographic projection with one ray per pixel and outputs a Surface, a pixel-art G-buffer containing normals, height, depth, and material IDs. This node sits at the center of the 3D → pixel-art lane for 2:1 isometric blocks and eight-direction characters.

• Does not perform lighting: downstream nodes such as サーフェース光プレビュー add light.

Usage tips

• The minimum setup is ソリッドボックス → ソリッドレンダリング → サーフェース光プレビュー. Start here to learn how a volume becomes pixel art.
• Set pixel density with pixels_per_unit: at 16, a 1×1×1 box is approximately 16px. Match asset size with this value, not by moving the camera.
• The fastest way to find an angle is to drag in the 3D viewer, then write it back with この角度を採用.
• Use AutoFit for one still image; always use FixedCanvas for animation, eight directions, or a turntable. Auto-fitting changes dimensions with orientation and makes multiple frames jitter.
• The flat image output is for checking coverage. The primary output is surface followed by lighting nodes.

Common pitfalls

• Frame dimensions or center shift between frames: rotation is still using frame_mode = AutoFit. Switch to FixedCanvas and make frame_world_* large enough for the subject in every frame.
• The screen is plain white/black with no depth: this node intentionally does not light. Connect surface to surface_light_preview or a lighting definition.
• Eight-direction angles are slightly off: set camera_angle_snap = 45 to round them and absorb floating-point drift during animation.
• The outline is blurry: this node cannot blur because it has no AA and uses point sampling. Check enlargement filtering in the display.

Technical details

• AutoFit fits output dimensions to the volume bounds, up to 1024px.
• FixedCanvas fixes output dimensions and look-at center; use it for turntables and eight-direction assets.
• One output pixel = one ray, with point sampling and no AA for pixel art.
• normal_map is view-space: x right, y down in the image, z toward the viewer. This matches the サーフェース光プレビュー contract.
• height_map is normalized world z, white = high; depth_map is view depth, white = near.
• curvature_map is screen-space curvature: 0.5 = flat, >0.5 = convex edge. It is used for edge highlights.
• face_ids use {solid_id}:top / :left / :right / :front / :back / :bottom. If a generator's solid_id is blank, an automatic ID such as solid:0 is used.
• Implementation is crates/pxn-core/src/sdf_render.rs with rayon parallelism. Reference performance: about 3.3ms for 129×81px with 20 primitives in release mode.

3D viewer integration

• The 3D view in the inspector or preview window ray-marches the connected Solid input directly in Solid3DStage.
• Drag in the viewer to find an angle, then press この角度を採用 to write back azimuth, elevation_preset=Custom, and custom_elevation.
• The viewer checker and light gizmo are display-only and are not baked into surface or image output.

Examples

• ソリッドボックス → ソリッドレンダリング → サーフェース光プレビュー is the minimum isometric-block setup.
• TimeSource → SawtoothWave → ソリッド変換.回転Z → ソリッドレンダリング.image → AnimationRender creates a turntable sprite sheet.
• For eight directions, use VariantRange(key=render:azimuth, start=0, end=315, count=8) → BatchRender.items together with camera_angle_snap=45 and FixedCanvas.
• Connect サーフェース書き出し to produce a map set for an HD-2D engine.

Recommendations for animation

• frame_mode = FixedCanvas
• Set fixed_canvas_width / fixed_canvas_height to final single-frame dimensions.
• Make frame_world_width/depth/height large enough to contain the subject in every frame.
• For eight-direction assets, use camera_angle_snap = 45.

Related nodes

• solid_box / solid_sphere / solid_combine — build the volume to photograph
• surface_light_preview — the go-to next step for lighting the output Surface
• model_render — equivalent camera for the Model mesh lane with shared parameter vocabulary
• surface_export — export the Surface as a game-engine texture set