OpenAI, the developer of ChatGPT, has revealed its new Sora project.
Sora is Japan and means sky. OpenAI
In December 2022, ChatGPT exploded the possibilities of artificial intelligence into the consciousness of the general public. During 2023, the whole world was talking about nothing but the possibilities of artificial intelligence in coding, making art, doing background research and managing work routines.
ChatGPT has also been joined by other phenomenal artificial intelligences, such as Midjourney, which takes accurate pictures, and Dall-E, which produces fast art blasts in an instant.
However, they are still small compared to what OpenAI has now put on the table.
The company has announced that it will develop Gravel-artificial intelligence model, which is capable of creating videos of up to one minute long based on the inputs given by the user.
The story continues below the video.
TechCrunch says that Sora’s videos are in 1080p resolution, and they take into account movement, lighting, the environment as well as numerous different characters. Sora can also supplement existing video clips, meaning it will do its best to fill in missing data from videos. In other words, it can fix video artifacts and increase video frame rate.
All this is handled by text input, i.e. the user simply tells Sora what he wants Sora to do.
On Sora’s website, OpenAI presents the capabilities of its artificial intelligence with a really amazing video, from which the main image of this article was also taken.
That video was created with a feed “A stylish woman walks down a Tokyo street filled with warm glowing neon and animated city signage. She wears a black leather jacket, a long red dress, and black boots, and carries a black purse. She wears sunglasses and red lipstick. She walks confidently and casually. The street is damp and reflective, creating a mirror effect of the colorful lights. Many pedestrians walk about”i.e. the video was created by chatting to a wide range of languages.
Videos can be created photo-realistically, but also in different styles, such as animated and black and white. You can also use a photo to create videos and prompt Sora to continue from there.
The Verge says that Sora is currently only open to a limited user base, who evaluate Sora in terms of its potential risks, such as potential for abuse. OpenAI also accepts feedback.
The potential for misuse in terms of disinformation is admittedly huge. The New York Times says that OpenAI stamps all videos made by Sora with a kind of watermark that reveals the video is the creation of artificial intelligence. However, OpenAI recognizes that these watermarks can also be removed.
OpenAI does not disclose which videos and from which sources it has used as Sora’s learning material. OpenAI only says that the AI model has been trained on publicly available videos and copyrighted material that has been separately licensed.
Either way, the rate of development of artificial intelligence is amazing. by Sam Sheffer message in X illustrates how wildly text-to-video artificial intelligence has developed in just one year.
OpenAI itself admits that Sora may sometimes have difficulties creating videos. It’s not surprising. Sometimes videos have artifacts or their characters move weirdly. Sometimes the results are pretty acid trips.