AI Model Wiki

    Skip to main content
    45 min
    Lesson 1 of 5

    Introduction to Luma AI & 3D Capture

    From the course Luma AI: 3D & Video from Photos

    Luma AI brings photorealistic 3D capture and AI video generation to creators, developers, and businesses through accessible, browser-based tools.

    Luma AI Products

    3D Capture (NeRF)

    • Capture real-world objects and scenes in photorealistic 3D
    • Uses Neural Radiance Fields (NeRF) technology
    • Works from photos or video taken on any smartphone
    • Export for web, games, AR/VR, and e-commerce

    Dream Machine

    • AI video generation from text prompts and images
    • High-quality, cinematic video output
    • Fast generation times (120 frames in ~2 minutes)
    • Camera control and motion guidance

    3D Capture Workflow

    1. Capture: Walk around an object or scene recording video on your phone
    2. Upload: Send the video to Luma AI for processing
    3. Process: AI reconstructs the 3D scene using NeRF
    4. View: Explore the 3D capture from any angle in your browser
    5. Export: Download as GLB, OBJ, or embed on websites

    Capture Best Practices

    For Objects

    • Walk in a complete circle around the object
    • Capture from multiple heights (eye level, above, below)
    • Maintain steady, slow movement
    • Ensure consistent lighting throughout
    • Include 50-100 frames for best quality

    For Scenes

    • Walk through the space methodically
    • Cover all areas you want captured
    • Avoid reflective surfaces and transparent materials
    • Keep consistent exposure (avoid auto-exposure changes)

    NeRF Technology Explained

    Neural Radiance Fields represent 3D scenes as continuous volumetric functions:

    • The AI learns the color and density at every point in 3D space
    • Novel viewpoints are rendered by casting rays through the learned volume
    • Results are photorealistic, capturing reflections, transparency, and fine detail
    • Far superior to traditional photogrammetry for most scenes

    Named Example: IKEA & Snap AR

    IKEA's product visualization team uses Luma AI's 3D capture to create product assets for their Place AR app, enabling customers to see furniture in their homes before buying. Snap (Snapchat) integrated Luma's NeRF capture technology into several branded AR Lens experiences, allowing users to scan a physical product and place it in their environment as a 3D AR object.

    Common Pitfall: Capturing shiny, reflective, or transparent objects (glass bottles, chrome fixtures, clear packaging) produces poor NeRF results because the AI cannot reliably model the view-dependent reflections. For these objects, use HDRI-lit white background photography with photogrammetry as an alternative.

    Best Practice: For object captures, use a lazy Susan turntable and capture in two passes — one at eye level and one from a slightly elevated angle. This systematic approach guarantees coverage of the top surfaces that a freehand walk-around often misses, significantly improving reconstruction completeness.

    Key Takeaways

    • •Luma AI offers 3D capture via NeRF technology and AI video generation via Dream Machine
    • •3D captures created from smartphone video produce photorealistic, explorable 3D scenes
    • •Steady, comprehensive capture from multiple angles and heights ensures best reconstruction quality
    • •NeRF technology surpasses traditional photogrammetry with photorealistic volumetric rendering
    0 of 5 lessons marked complete

    Progress is kept in this browser only.