Sora and Text-to-Video Quiz

Answers and explanations appear as you take the quiz.

1. What core capability defines a text-to-video model like Sora?
   • It generates a video clip from a written text prompt
   • It edits audio tracks
   • It compresses existing movies
   • It translates subtitles between languages
2. What is the central technical challenge that makes video generation harder than image generation?
   • Videos use more colors
   • Maintaining consistency and coherent motion across many frames over time
   • Videos cannot be stored digitally
   • Text prompts are illegal for video
3. How does Sora represent video internally to feed its transformer?
   • As one giant pixel grid per clip
   • As audio waveforms
   • As small space-time patches that act like tokens
   • As handwritten frames
4. Which generative technique underlies Sora's frame synthesis?
   • Rule-based animation scripts
   • Simple frame copying
   • Manual rotoscoping
   • Diffusion, gradually denoising random noise into a coherent clip
5. Which limitation did OpenAI’s 2024 Sora report document?
   • Physics and object consistency can break, like morphing objects or extra fingers
   • They refuse to use color
   • They can only make black-and-white clips
   • They always produce silent films on purpose
6. Roughly how long are the clips Sora was demonstrated generating at its 2024 reveal?
   • Several hours
   • Up to about a minute
   • Exactly 24 hours
   • Only one frame
7. What joint training data does OpenAI’s 2024 Sora report describe?
   • Only paired audio and transcripts
   • Only text documents
   • Video and image data
   • Only 3D motion-capture skeletons
8. What input can the 2024 Sora report condition on in addition to text?
   • Only a written script with fixed timestamps
   • A spreadsheet of camera exposure settings only
   • A sound waveform with no visual prompt
   • An image that the model can animate into a video