OpenAI’s AI video generator Sora to launch later this year: report

OpenAI reportedly plans to launch its latest generative artificial intelligence (AI) model, Sora, later this year. Launched last month, Sora is an AI-powered text-to-video generator that the company claims can generate videos up to 60 seconds long, much longer than any industry competitor. In a recent interview, the AI ​​company’s CTO also revealed that Sora can generate videos in minutes depending on the complexity of the prompt. The interview also covered topics such as the model’s flaws, the data used to train the model, and Sora’s possible price point.

in a interview In an interview with the Wall Street Journal, Mira Murati, chief technology officer of OpenAI, said when talking about when the artificial intelligence video generator will be launched, “I definitely hope to launch it this year, but it may take a few months.” When Asked whether the U.S. presidential election, slated for November 2024, would affect when the company releases its AI models, Murati said it was important for the company to address misinformation and harmful bias, adding: “We will not release any Content that we have no confidence in how it might impact elections around the world.”

Mulati was also asked about the source of data on Sola’s training. The issue is also of interest to Italy’s data protection regulator, which is investigating OpenAI for possible data breaches. The regulator gave the AI ​​company 20 days to provide information about the AI-powered video generator.

In the interview, Mulati also didn’t seem to reveal any details, saying only that the AI ​​model was trained on data available in the public domain and that the data was licensed to the company. She also said she wasn’t sure if data from YouTube or Facebook was used, but confirmed to the publication that Sora used content from Shutterstock, with whom they have a partnership.

Murati explained that Sora is also very expensive to run compared to other OpenAI tools like ChatGPT and DALL-E. “ChatGPT and DALL-E are optimized for public use, while Sora is actually a research effort. It’s much more expensive,” the CTO told the Wall Street Journal. While the company has not yet finalized the price point for the AI ​​video model, Murati said the company is trying to offer the model at a similar cost to the DALL-E once it is released. Currently, the base pricing for the DALL-E 3 model is $0.040 (approximately Rs. 3) per image with a resolution of 1,024 x 1,024 pixels.

Additionally, the OpenAI CTO emphasized that the company is working hard to bring audio capabilities to Sora, optimize generation times and its cost-effectiveness, and make further improvements to minimize defects.