OpenAI surprises again: here is Sora, the AI that creates photorealistic videos
Another interesting project from OpenAI. The company, in fact, launched its first text-to-video model, i.e. artificial intelligence Sora. The model, which is already reaching the first content creators, is still in an experimental phase, however it is already possible to view the first videos created with this system.
The final results they are quite appreciable given that the scenes, reconstructed using AI starting from textual input, are realistic and credible. In this sense, the company has specified that Sora is capable of creating “complex scenes with multiple characters, specific types of movement, and fine subject and background detail.” Also, the model can also understand how objects “exist in the physical world”beyond “accurately interpret props and generate compelling characters that express vibrant emotions.”
Still, Sora is able to generate a movie also starting from a static image as well as adding missing frames (it will also be possible to extend a video).
The final results, as already mentioned previously, are at times impressive, although there is no shortage of uncertainties. In this regard, OpenAI itself highlighted that the model “may have difficulty accurately simulating the physics of a complex scene”. At least for the moment, the maximum length of videos is one minute.
