The small LiDAR sensor on the back of iPhone and iPad Pros has often been overlooked, yet it holds far greater potential than most people realize. KIRI is excited to announce KIRI Engine Version 3.14, which centers on a complete revamp of LiDAR Scene Scan. This update transforms the previously limited 3D visualizations into high-fidelity, immersive models that accurately capture and restore virtually any environment. LiDAR Scan on iPhone Pro and iPad Pro, source from online. What is LiDAR Scene Scan? Alongside LiDAR Room Scan and Object Capture, LiDAR Scene Scan has been a valuable tool for KIRI Engine users with LiDAR-equipped iOS devices, enabling fast environment visualizations with generalized areas and colors. At its core, this technique combines depth data captured from the LiDAR sensor with color data from the camera. The process is relatively quick and easy, making it a practical choice for basic scene restoration. However, the method has two major limitations: Lack of detail and difficulty with complex surfaces: While effective for capturing the general structure of a space, the technique struggles to accurately represent finer details, especially on reflective, transparent, or translucent surfaces. This lack of precision limits its viability for professional uses such as surveying, architectural scouting, and interior design. Overshadowed by emerging capture methods: With rapid advancements in 3D reconstruction, newer technologies have surpassed traditional scene scans by delivering higher levels of detail and full-scene accuracy. Notable examples include KIRI’s Photo Scan, which produces game-ready high-quality 3D assets, and 3DGS, which enables the capture of massive environments with realistic rendering. Capturing with LiDAR Scan. Although various methods have addressed certain limitations, a gap remains: a solution that combines direct depth data capture with high-quality scene restoration. This is precisely where KIRI Engine’s optimized and enhanced Scene Scan sets a new standard. AI-Enhanced LiDAR Scene Scan in KIRI Engine 3.14 While many platforms utilize the LiDAR sensors on select iOS devices, KIRI recognized a critical limitation: direct LiDAR scene scans often produce low-resolution, incomplete outputs that fail to capture the true richness of physical environments. Common iPhone/iPad LiDAR Scan Results. Determined to unlock the full potential of the depth and visual data available, KIRI Engine 3.14 introduces AI-Enhanced Scene Scan — a major advancement that leverages cutting-edge machine learning and dedicated algorithms to refine both surface orientation and geometric depth information. Before exploring the details of this update, it is important to first understand the fundamental principles of LiDAR scanning. LiDAR, which stands for Light Detection and Ranging, measures depth data to generate 3D maps of objects and environments. It operates similarly to radar systems, which send out sound waves and measure the time it takes for them to reflect back; however, LiDAR uses light instead. This method allows for incredibly fast, detailed, and versatile spatial mapping across a wide range of applications, from topographical surveys to autonomous vehicles. Although professional-grade LiDAR scanners are typically very costly, Apple has successfully integrated these intricate sensors into iPhone and iPad Pros. This integration enables direct depth capture and, more importantly in KIRI’s case, allows for efficient and highly accurate 3D capturing and processing. In practice, LiDAR scanning is quick and efficient, but it remains limited by its reliance on depth maps. Due to the nature of light-based measurement, LiDAR sensors often struggle with reflective, transparent, or translucent surfaces: a key challenge that KIRI Engine 3.14 aims to overcome. Same human captured and reconstructed by four different LiDAR-based programs. At the heart of this enhancement are two specialized machine learning models, each targeting a different aspect of 3D scene reconstruction. First, StableNormal enhances surface normal estimation directly from image inputs. Traditional methods often produce unstable or blurry surface details, especially under poor lighting, motion blur, or when dealing with reflective materials. StableNormal solves this by using a two-step process: it first creates a strong initial prediction of surface directions, then carefully refines the details to sharpen the geometry. The result is a highly stable and accurate normal map that captures fine surface structures even in difficult scanning conditions. StableNormal Normal Estimation and Applications. Second, PromptDA refines the depth data captured by the LiDAR sensor. Rather than relying solely on raw, low-resolution LiDAR measurements, PromptDA uses the LiDAR signal as a prompt to guide a powerful depth foundation model, drastically improving the resolution, density, and metric accuracy of the resulting depth maps. This fusion process results in highly detailed, high-fidelity geometric structures that significantly outperform traditional raw iPhone-based depth captures. PromptDA Depth Estimation Comparison (against raw LiDAR depth data). By combining enhanced surface normal maps with refined, metric-accurate depth data, KIRI Engine’s AI-Enhanced Scene Scan sets a new standard for mobile 3D scene reconstruction, delivering outputs that are both visually coherent and structurally precise. Same environment captured and reconstructed by four different LiDAR-based programs. The image above compares scans of the same environment, captured along identical paths and angles but processed through different programs. Programs A, B, and C struggle with overall surface reconstruction, producing uneven geometry, distorted shelving, and heavy noise across structural elements like boxes, floor planes, and walls. Fine object shapes are flattened or lost entirely. In contrast, KIRI Engine’s AI-Enhanced Scene Scan preserves both structural clarity and fine surface detail with far greater accuracy. You can actually see the individual shapes of the water bottles on the rectangular rack, where other programs failed to distinguish them due to the transparent nature of the material, instead rendering them as indistinct blobs. Using AI-Enhanced Scene Scan The Scene Scan capturing process remains simple and intuitive. What has changed is the addition of an optional cloud-based enhancement step, bringing cutting-edge 3D research techniques directly to users. In addition to improved scan quality, KIRI Engine now offers the ability to export Scene Scans as raw datasets with full camera metadata, formatted for 3D research projects like nerfstudio, GauStudio, and others. Each dataset includes the original photos, confidence maps, depth maps, and camera poses, opening new opportunities for academic, research, and professional projects. Uploading LiDAR Scene Scan for Enhanced Cloud Processing. To unlock these features, users must meet two prerequisites: they must use a LiDAR-equipped iOS device and have an active KIRI Engine Pro subscription. The raw data export is exclusive to Scene Scans and is offered completely free of charge. Users can enable it by toggling Developer Mode within the app under ‘Me’ → ‘Developer Mode.’ As iPhone and iPad Pros become more widely adopted compared to traditional LiDAR scanners, KIRI aims to empower a new generation of academics, researchers, and creators to explore and expand the frontiers of 3D scanning. Looking Ahead As the 3D research field continues to advance, KIRI Engine is committed to bringing state-of-the-art innovations to users and expanding the possibilities of 3D reconstruction technology. This update draws inspiration from the groundbreaking work behind AGS-Mesh, PromptDA, and StableNormal, whose contributions have helped push the boundaries of what is possible. Moving forward, KIRI aims not only to refine 3D reconstruction as a tool but also to broaden how we connect with, understand, and reshape the world around us. Brought to you by KIRI Innovations: This article is part of the befores & afters VFX Insight series. If you’d like to promote your VFX/animation/CG tech or service, you can find out more about the VFX Insight series here. The post KIRI Engine’s New iPhone Update Uses AI To Fix LiDAR’s Quality Limitations appeared first on befores & afters.