After announcing the arrival of a permanent memory function on ChatGPT, OpenAI lifted the veil on February 15, 2024 on a new artificial intelligence model capable of generate videos “realistic and imaginative†using a simple text query: Sora.
Sam Altam’s company indicates that its application allows the creation of “Complex scenes with multiple characters, specific movement types, and precise subject and background details†, lasting a maximum of one minute.
Sora can provide multiple shots within a single generated video, all “faithfully retaining the characters and visual style†. The tool also allows you to create a video based on an image or to complete an existing video.
On the page dedicated to Sora, we learn that the AI relies on a very advanced interpretation system to generate its videos. OpenAI says its new model can understand how objects “exist in the physical world†, “accurately interpret messages and create compelling characters that express vibrant emotions†.
At the moment, Sora is only accessible to a limited group of testers. The company specifies working with “experts in areas such as misinformation, hateful content and bias†, but also visual artists, designers and filmmakers. The objective is to collect as much external feedback as possible in order to refine the tool before its deployment to the general public, whether for the sake of security or quality. .
OpenAI still mentions some weaknesses inherent to its AI model for converting text into video: “He may struggle to accurately simulate the physics of a complex scene and may not understand specific cases of cause and effect†.