Making films with "magic"
OpenAI has built a new video model called Sora. It is a new tool that can make a video like magic! Just tell Sora your imagination, and watch it create high-quality video up to one minute just in front of your eyes.
How good is it?
Videos created by Sora look good on both big and small devices(设备), such as computers and smartphones. Sora can also make scenes with lots of people doing different things with clear details. It also knows things should look in real life.
Why is it so good?
Sora is a good learner. It looks at many videos and learns from them by (break) them into very small bits. It uses these bits to make a new video. The new video is grey and a complete mess at first—you can't tell what's in it. Then, Sora fixes the video until it looks good and smooth.
Sora also gets help from the AI chatbot, which turns (user) short sentences into clear instructions. This helps Sora make the video just as it (ask).
Is it good enough?
Since Sora was announced in February 2024, OpenAI (know) that it is far from perfect. When meeting complex images (复杂图像), it can't make everything look real because it may find them hard (understand). For example, when a person moves, things around them change. In another video, a grandma blows the candles on a birthday cake, but the flame(火焰) (not) move at all. Sora may not be able to tell "left" from "right" (correct) and mess up a person's leg moves.