Today, the same day Google chose to announce Gemini 1.5, OpenAI decided to counter the program by introducing Soraa new generative model (a specialty of the house) that once again surprises us with what it can create from the simple call text. And yes, before I get into the news, the eternal clarification that the result is not really perfect, that it does not take much effort to detect inconsistencies and errors. But even so, it is true that his creations are more than remarkable.

Sora, as you can see in the videos below, allows you to create short videos (under one minute) with different styles, from animation to realistic appearance, from the challenge in which we determine its content and format. Let’s look at three examples shown by OpenAI.

call: Several giant mammoths approach across a snowy meadow, their long woolly fur moving slightly in the wind as they walk, in the distance are snow-covered trees and magnificent snow-capped mountains, in the afternoon light with soft clouds and the sun high in the distance creating a warm glow, low camera view it is stunning and captures a large furry mammal with beautiful photography and depth of field.

Result

call: Movie trailer featuring the adventures of a 30-year-old spaceman with a red woolen knitted motorcycle helmet, blue sky, salt desert, film style, shot on 35mm film, vivid colors.

Result

call: A drone view of the waves crashing against the steep cliffs along Garay Point Beach in Big Sur. The crashing blue waters create white waves while the golden light of the setting sun illuminates the rocky shore. In the distance is a small island with a lighthouse and green bushes cover the edge of the cliff. The steep drop from the road to the beach is a spectacular feat, with cliff edges jutting out over the sea. This is a view that captures the raw beauty of the coast and the rugged landscape of the Pacific Coast Highway.

Result

These are just three of the 48 examples published by OpenAI on the website where Sora presented. Of course, if you’re interested in trying it out, I’m afraid you’ll have to wait a bit longer, as the company has decided not to offer it to the public just yet, as it wants to test it first to ensure it’s safe and also to minimize the risk of them being used to create false content intended to misinform.

For this purpose, as they specify on their website, several filters are set up. Therefore, if Sora finds that the entry (challenge) violates the policy of use of the service (extreme violence, sexual content, hateful images, images of celebrities or third-party intellectual property), the creation request will be rejected. Except, OpenAI is also working on implementing a watermarking system in videos as well as recently in DALL-E.