There are 3 ways to create videos with Sora: text-to-video, image-to-video, and video-to-video. Next, we will demonstrate these three operation methods respectively.
1. Text-to-video
Generate a video by entering a prompt. You do not need to enter complex commands as you would with Stable Diffusion or Midjourney; here you only need to enter natural language. This includes cinematography methods and camera movement methods, all implemented directly using natural language. Suppose you are a director; then you can directly upload your script storyboard, which is the easiest. The entire input process supports Chinese. Here we provide an example.
- Scene: African savanna, hot summer
- Shot: A sparrow (character) weaving through the jungle
- Style: Documentary
- Lighting: In the afternoon sunlight there are wisps of clouds; the sun hangs high in the sky, casting warm rays from afar
- Camera movement: Close-up of the bird flying
If you want to use the bird as a character to create a coherent and consistent bird, then you need to describe the character strictly so that it remains consistent when reused. We will cover this issue in another video.
Although this seems like a simple example, there is actually a lot of knowledge in it. We will break down these five elements separately so that you can learn to precisely control the video.
2. Image-to-video
Click image to video, then click the image button to upload an image. First, have Sora analyze the image to ensure it understands it correctly. Then you can ask Sora to complete the specified action. For example, here we ask it to make the dog in the image sway left and right. This is a simple example; you can also describe detailed shot content as in the text-to-video case.
3. Video-to-video
For video-to-video, click video to viideo and upload a video. After a successful upload, it will analyze it. At this point, you can perform multiple operations, such as optimizing, modifying, and extending. |