After having tested the existence of a ChatGPT based search engine, here's something new from OpenAI. This time officially Sora debuts as the new OpenAI product that generate videos starting from text. The model, so-called text-to-video, also paves the way for innovative solutions for creative professionals and researchers and not just for users who will be able to use them in the future.
Sora by OpenAI: where imagination and reality meet
This text-video template is designed for creating video scenes up to one minute in length that maintain high visual quality and are faithful to user indications. What makes Sora particularly compelling is his ability to generate complex scenarios with multiple characters, specific movement patterns, and accurate detail on subjects and backgrounds, demonstrating a deep understanding of real-world language and physical dynamics.
Despite the innovativeness, Sora is not without challenges, such as difficulty correctly simulating the physics of complex scenes or managing precise spatial details.
The implementation of Sora in OpenAI tools will also build on security methodologies already developed for products such as DALL E 3, ensuring that the content generated complies with strict policies. Furthermore, the model benefits from learning techniques deriving from research on DALL·E e GPT language models, such as the recaptioning technique, which allows you to faithfully follow the textual instructions in the video output.
Sora's ability to generate videos from text instructions or extend existing videos opens up new perspectives for content creation. This model not only supports the creative community, but lays the foundation for future developments in artificial intelligence, bringing us closer to the goal ofgeneral artificial intelligence (AGI).
OpenAI has not yet announced when Free, Plus, or Enterprise users will be able to use Sora. The model is currently under development although the first results are already visible on OpenAI official website.