Do you remember those videos made with AI that were fuel for your worst nightmares? Artificial intelligence companies have been refining their models over the past year, releasing better chatbots and image and audio generators, but there was one pending task: video. Now OpenAI is back to bring us a new AI that will cause quite a stir.
The company behind the successful ChatGPT has launched Sora, its new video generation model, capable of “creating realistic and imaginative scenes from text instructions”. Thanks to this new AI tool, users can create photorealistic videos based on the prompts they provide. However, the maximum duration of the generated videos is only one minute.
According to OpenAI details on their website, Sora is capable of creating “complex scenes with multiple characters”, specific types of movement, and precise details of the subject and background, as well as understanding how objects “exist in the physical world” and “accurately interpreting props and generating convincing characters that express vibrant emotions”.
To top it off, this new model is also capable of generating a video from a still image, as well as filling in missing frames in an existing video or extending it. Among the examples provided by OpenAI, we can see an aerial scene of California during the gold rush and a video that appears to be shot from inside a train in Tokyo.
Yes, as many may observe, AI is still not perfect and “AI elements” can be seen in the generated videos, such as objects that appear out of nowhere or people moving in a strange way. The thing is, even with all those flaws, Sora is currently the most accurate video generating model in the industry, something that could be a headache for organizations fighting against online disinformation.
For now, access to Sora is limited, being only available to some visual artists, designers, and filmmakers, as well as to OpenAI’s own “red teams” responsible for evaluating the model in order to detect possible harms and risks.
