루마 AI 퀴즈

퀴즈를 풀면 정답과 설명이 나타납니다.

1. What did Luma’s historical Dream Machine product generate?
   • Short video clips conditioned on text or image inputs
   • Three-dimensional CAD files from audio alone
   • A guarantee of realistic physics in every clip
   • Neural radiance fields from a single unposed image
2. Which research approach reconstructs a 3D scene from images by representing it as a continuous radiance field?
   • CTC sequence alignment
   • Neural radiance fields (NeRF)
   • Region proposal networks
   • 그라데이션 부스팅
3. 단일 이미지가 아닌 비디오를 생성할 때 가장 어려운 기술적 과제는 무엇입니까?
   • 글꼴 선택
   • 단일 JPEG 압축
   • 프레임 전반에 걸쳐 시간적 일관성 유지
   • 도메인 이름 선택
4. What did Luma introduce as Camera Motion Concepts for its Ray2 video model?
   • A guarantee that generated videos match real footage
   • A NeRF feature that requires no training examples
   • A rule that keeps every object fixed in every frame
   • Learnable camera moves that can be composed from one or a few examples
5. What input does the original NeRF paper use to learn a radiance field for a scene?
   • Multiple posed images of the scene
   • A single text prompt with no images
   • An unordered list of object names
   • A video transcript with no camera information
6. What does a NeRF query at a 3D location and viewing direction?
   • A tokenized transcript and speaker label
   • A color and volume density used for rendering
   • A camera-motion command and frame index
   • A bounding box and object class only
7. Which video model does Luma’s July 2026 official information page identify as current?
   • Dream Machine, which remains the current model
   • Ray2, which is not deprecated
   • Ray3.2
   • NeRF, a product for text-to-video generation
8. What should a user check before treating a generated clip as factual footage?
   • Whether the file uses a common codec
   • Whether it has a high frame rate
   • Whether it contains a title card
   • Whether the clip is synthetic and what prompt or source material produced it