Chapter 3 — Settings

WHERE

Settings → General → Interface → Default Mode picker. Default: "Simple Mode".

TECHNICAL

Controls which of two UI modes the app opens in after the next launch. "Simple Mode" is the guided wizard workflow with 4 steps (Import → Processing → Preview → Export, documented in Chapter 10 under Z1–Z4), "Expert Mode" is the classic three-panel layout with Navigator, 3D viewport, and Expert Inspector from Chapter 2. The value is remembered across restarts. Same effect as the menu Mode → Simple Mode (⌘1) / Mode → Expert Mode (⌘2), except that the menu switches the current session while this picker sets the default for future sessions. Both modes access the same project state — projects, cameras, and training configuration are preserved when switching modes. Mode-specific toolbar buttons are re-rendered immediately.

WHERE

Settings → General → Interface → Toggle "Pause after frame extraction (Expert)". Default: off.

TECHNICAL

Applies exclusively to Expert Mode and there exclusively to video footage. When the toggle is off (default), the large button reads "Start Processing" and the run goes through in one pass: extract frames from the video, check image quality, align cameras, train. When it's on, the same button reads "Extract Frames" and the run halts after extracting and reviewing the individual frames. During this pause you can review the image set in the Navigator, remove individual images, and change the training settings before starting the rest yourself. The value is remembered across restarts. It has no effect on Simple Mode or on imports that already consist of individual images — there's nothing to extract there.

WHERE

Settings → General → Interface → Toggle "Show marker recipe suggestions". Default: on.

TECHNICAL

Allows the app to silently search your images for AprilTag markers after an import and, if applicable, suggest a different SfM recipe. The notice appears as a banner in the Expert Inspector above the Cameras & Capture section and offers to switch to Professional, which uses repeatedly detected markers as rigid anchors for more accurate camera positions. Only what's actually usable gets counted: a marker ID must appear in at least three images, otherwise it doesn't yield a usable track and remains uncounted — so several different IDs that each appear only once trigger nothing. The notice also only appears if the native SfM is actually running and no higher-value recipe is set yet, and it disappears as soon as a run starts. The toggle is the permanent form of "Don't suggest again" in the banner itself; clicking that switches this off. Conversely, re-enabling it here counts as an explicit fresh start: even a notice dismissed for the current import is checked again.

WHERE

Settings → General → Interface → Language picker. Default: "System Default" (follows the macOS language).

TECHNICAL

Selects the display language of the entire app UI, independent of the macOS system language. The picker lists the entry "System Default" and 18 fixed languages: English, Deutsch, Français, Español, Português, Italiano, Nederlands, 日本語, 中文, 한국어, Svenska, Norsk, Dansk, Suomi, Polski, Čeština, Русский, Türkçe. With "System Default", the app follows the macOS language. With an explicit choice, the language setting is remembered across restarts; full effect generally requires an app restart, because localization bundles are only loaded at launch.

WHERE

Settings → General → Viewport → Background picker. Default: "Dark Gray" (RGB 0.1, 0.1, 0.1).

TECHNICAL

Sets the background color of the 3D viewport. Three options: "Dark Gray" (RGB 0.1, 0.1, 0.1 — default), "Black" (0, 0, 0) and "White" (1, 1, 1). The choice takes effect immediately on the running Metal renderer and is stored in the User Defaults, so it survives app restarts. There is exactly one setting with three controls: this picker, the menu Viewport → Background (M21, M22, M23), and the B key in the viewport all write through the same setter — whoever toggles it via menu or key will see the new value reflected here in the Settings picker afterward. (Up through v1.7 the app didn't remember the choice, and the word "Default" in the picker was somewhat dishonest in that regard; since 2026-07-18 the three paths are consistent.) A deliberate consequence of this change: a white background stays white even after a restart, and thus also ends up in screenshots and orbit video exports. Still valid for recordings: white backgrounds highlight green/blue floaters more strongly, dark backgrounds are better for clean render captures.

WHERE

Settings → General → Viewport → Color picker "Selection color". Default: red (RGB 1, 0, 0).

TECHNICAL

Sets the color used to tint Gaussian points selected in edit mode — i.e. what you've just marked with the brush or via the clipping plane and would delete or desaturate in the next step. The picker doesn't support transparency. A change is written to the User Defaults immediately and passed on to the running renderer, so the view recolors with the next frame; a restart is not required. The tint is display-only and appears in no exported file.

WHERE

Settings → General → Viewport → Color picker "Cylinder color". Default: cyan (RGB 0.1, 0.9, 1.0).

TECHNICAL

Sets the color of the cylinder crop — both the wireframe and the content the cylinder encloses. Careful, the two tones mean opposites: the selection color (S39) marks what is to be removed, the cylinder color marks what remains. That's why they're configurable separately and clearly distinct out of the box. Like S39, no transparency, takes effect immediately, display-only.

WHERE

Settings → General → Viewport → Toggle "Invert vertical navigation". Default: off.

TECHNICAL

Reverses the up/down direction of viewport navigation. Scope since 2026-07-18: orbiting, look direction in fly mode, and panning — the latter both via gesture and when dragging with the right mouse button, which previously didn't read the flag at all. Before, the toggle only worked half-way; anyone who enabled it moved two gestures against each other. When panning, vertical mouse movement inherently has the opposite sign compared to orbiting, which is why it's conditionally inverted there instead of copied from orbiting — anyone who never touches the toggle won't notice the correction. Zoom is deliberately excluded: there, the system's scroll direction already determines the sign, and the deltas control distance and viewing angle, not vertical navigation.

WHERE

Settings → General → Training → Live Preview picker. Default: "Every 50 Iterations".

TECHNICAL

Determines at what iteration interval the running training snapshot is rendered into the 3D viewport. Four discrete values: 0 ("Off"), 50, 250, 1000 iterations. With live preview active, the trainer copies the Gaussian buffer from the GPU into a separate render buffer and triggers a viewport redraw. With "Off", the viewport is only updated after training completes. Performance cost: every 50 iterations ~5–10% slower on M3 Ultra, every 250 iterations ~1–2% slower, every 1000 iterations unmeasurable. Memory overhead constant ~2 GB for the snapshot buffer, independent of the interval. Scope: the value belongs to the app, not the scene. It is carried along with the training settings for technical reasons, but is explicitly preserved when opening a scene file, when switching the preset, and during automatic 360° switching — previously an opened scene would overwrite its own preview rate with that of the scene's author, "Off" included. For the same reason it doesn't count toward the recipe comparison and doesn't trigger a "Modified" badge on the preset. "Continue Training" also follows the current value rather than the one that was in effect during the original run — since then a continuation shows a preview at all and honors it if you turned it off in the meantime. It is also the only setting in this window that doesn't survive an app restart: unlike every other entry here, it is not stored permanently anywhere; every launch starts again at "Every 50 Iterations". A training run already in progress does not change its rate anymore; a change takes effect on the next run. At interval 50 the visual impression is a smooth "growing" of the point cloud, at 1000 it looks jerky.

WHERE

Settings → General → Training → Toggle "Save a PLY automatically after each run". Default: on.

TECHNICAL

Writes the result as a PLY to the Exports folder after every completed run — fresh run and continuation — (training_<timestamp>.ply). The feature itself has existed since v1.7.0, but ran without any indication and without a toggle; since 2026-07-18 it is visible and can be switched off. Caution when disabling: the automatic write is a safety net — without it, a completed run exists only in memory until you export it yourself or save it as a scene. Caution when leaving it on: a file reaches ~2 GB for detail-rich scenes. If the setting was never touched, "on" applies: a missing value in the User Defaults is explicitly read as "on", not as "off" — otherwise every existing user would have silently lost the safety net. The target folder depends on the build: without sandboxing ~/Documents/RadianceKit/Exports, in the App Store version instead ~/Library/Containers/…/Documents/RadianceKit/Exports — hence the button next to it (S37).

WHERE

Settings → General → Training → Button "Show Exports Folder…". Pure action, no stored value.

TECHNICAL

Opens the folder in Finder that the automatic PLYs from S36 are written to. Exists because the path differs depending on the build, and in the App Store version the files sit in the app's container, where nobody looks — which is exactly why they repeatedly seemed "lost". The button and the writer use the same path definition, so they cannot diverge. No stored state.

WHERE

Settings → AI Helpers → first section → toggle „AI Helpers enabled". Default: on.

TECHNICAL

Master switch over all AI Helpers features in the pipeline. When off, the import and SfM pipeline skips all ML-based preprocessing stages entirely — no Apple Vision call, no CoreML model load, no NPU wake-up. When on, the individual sub-options (S12–S13) are consulted. The value is remembered across restarts. Affects the following stages: (a) frame quality pre-check before SfM (S12), (b) loop closure detection (S13). Important: when off, all sub-options in the tab are disabled and visually greyed out. A footer note emphasizes that all AI Helpers run strictly on-device — no image upload, no cloud processing. The privacy guarantee comes from using exclusively the Apple Vision framework (locally on the Neural Engine) and CoreML models that live directly in the app bundle.

WHERE

Settings → AI Helpers → Available section → toggle „Frame quality check". Default: on.

TECHNICAL

Enables the frame quality screener, which analyzes every imported frame before the SfM call. Each frame is checked for motion blur, over- and underexposure, insufficient image structure, and an obscured lens area; the finding is additionally graded by severity (unremarkable, warning, severe). Frames without a finding pass straight through. Flagged frames trigger a modal confirmation dialog that lists every problematic frame with a thumbnail and reasoning and asks whether it should be removed — optionally keep all, remove all flagged ones, or remove a custom selection. Important: no automatic deletion — the dialog is always required, the user retains the final decision. When off, all frames are passed on to SfM unchecked. When the master (S11) is disabled, this toggle is visually greyed out and without effect.

WHERE

Settings → AI Helpers → Available section → toggle „Loop closure detection". Default: on.

TECHNICAL

Enables the Apple Vision feature-print-based loop closure detection. For each imported frame, a feature print is computed, a neural embedding of the image content. The feature prints are then compared pairwise; sufficiently similar pairs that are far enough apart in capture order (i.e. not direct neighbours) are identified as „loop closure candidates" and written to a sidecar JSONL file in the project folder. Informational only — the imported image sequence is not modified. Purpose: gives the SfM solver (especially COLMAP) a hint that these frames belong together in a cluster in 3D space. For native SfM, the sidecar information is currently documentary only; COLMAP uses the hints internally via a custom matches file (manual integration possible, not automatically linked). When off, no feature prints are generated. Visually greyed out when the master (S11) is disabled.