It’s been over a year since OpenAI announced Sora – its video and audio generation model. OpenAI is now introducing the first major update with Sora 2, which can generate more realistic and physically accurate videos, along with improved user control and editing capabilities.
Described as a “GPT 3.5-moment”, Sora 2 also brings improved sound output with synchronized dialogue and sound effects. Sora 2 is able to avoid common problems associated with prior video generation models, such as morphing objects and distorting physical reality to complete a prompt.
Generated videos will also follow multi-shot instructions and maintain consistency across different scenes. Sora 2 supports several video generation styles, including cinematic, realistic and anime.
OpenAI is also launching a standalone Sora iOS app with an invite-only system. Sora 2 features several safeguards, including restrictions on using public figures in video generation. OpenAI also confirmed that users can give permission to others to create “cameos” with their likeness, but that can be revoked by the user at any time in the future.
Sora 2’s initial rollout starts today and will be limited to the US and Canada. As for pricing, OpenAI confirmed that Sora 2 will be free initially with “generous limits” so users can explore its capabilities.
ChatGPT Pro subscribers are getting access to a higher-quality Sora 2 Pro model. OpenAI also confirmed its plans to release Sora 2’s API soon.
