• Sora can create minute-long photorealistic videos with complex scenes, multiple characters, and accurate details. • It can interpret physical objects, props, and emotions, showcasing a remarkable leap in AI technology.
• Aerial view of California during the gold rush • Indian railway station with bullet trains • Tajmahal construction process Timelapse
• Achievements: Filling missing frames, extending existing videos, and generating videos from a single image. • Limitations: Acknowledges challenges in accurately simulating complex scenes but showcases impressive overall results.
Sora competes with Midjourney, Runway, Pika, and Google's Lumiere.
• Currently accessible to "red teamers" evaluating potential harms and risks. Visual artists, designers, and filmmakers also providing feedback. • Challenges: Acknowledges limitations in simulating physics and interpreting cause and effect.
• Transparency: Watermarking DALL-E 3 for authenticity. • Challenges: Addressing concerns of fake, AI-generated videos being confused for genuine content.