OpenAI Introduces New Generation AI Video Generation Model: Sora

OpenAI’s latest AI video generation tool, Sora, is capable of transforming short textual descriptions into detailed high-definition movie clips. This technology is an extension of DALL-E 3. Sora combines diffusion models and transformer network techniques to enhance the model’s training capabilities for videos of different resolutions and orientations by processing “chunks” of video data. However, the tool is currently limited to sharing with third-party security testers and selected video creators and artists to gather feedback on how to make Sora as useful as possible.

OpenAI is cautious about the launch of Sora, primarily considering the potential misuse of AI-generated videos, such as in political propaganda or misinformation. To prevent such occurrences, OpenAI has implemented several security measures in Sora, including filters running on all prompts and reviewing frames of generated videos to block materials that violate OpenAI’s safety policies. Additionally, the company has added industry-standard C2PA labels, which are metadata tags indicating content source and authenticity, to the video outputs.

These security measures are part of a series of measures launched by OpenAI ahead of the 2024 global elections to mitigate the risks of information interference that AI technology may bring.