Soon, you will be able to experience OpenAI's highly anticipated text-to-video generator, Sora. OpenAI's Chief Technology Officer, Mira Murati, revealed that Sora will be officially launched "this year," with a specific timeframe "possibly a few months later."

As early as February this year, OpenAI showcased the powerful capabilities of Sora. This tool can generate incredibly realistic scenes based on text prompts, which is truly impressive. Although initially only available to visual artists, designers, and filmmakers, some of the videos generated by Sora have quietly appeared on platforms like X, attracting widespread attention.

In addition to being open to the public, Murati also revealed that OpenAI plans to eventually add audio functionality to Sora, making the generated scenes even more vivid and realistic. At the same time, the company hopes that users will be able to edit the video content generated by Sora, as AI tools are not always 100% accurate when creating images. "We are exploring how to utilize this technology to create a tool that people can edit and create with," said Murati.

However, when asked about the data used by OpenAI to train Sora, Murati did not disclose too many details, seemingly avoiding the question. She stated, "I won't reveal too much about the details of the data used, but they are publicly available or licensed." Murati also did not provide a clear answer regarding whether data from YouTube, Facebook, and Instagram videos were used.

Furthermore, Murati pointed out that the "operating cost of Sora is relatively high." In order to lower the barrier of entry, OpenAI is working to make this tool available at a cost similar to the company's AI text-to-image model, DALL-E, when it is released to the public.

Murati also mentioned that after its release, Sora may not be able to generate images of public figures, similar to the policy of DALL-E. Additionally, in order to distinguish the generated videos from real content, the videos will have watermarks.